Research On Fault Diagnosis Of Railway Freight Car Bearing Based On Deep Learning

Railway freight car is an important means of transportation of goods in China,rolling bearings are one of the key components of railway freight car,and its good operating condition is the basis for ensuring the safe operation of trains.According to statistics,among all the operating faults of trains,the proportion of faults caused by bearing failure is about one-third.Compared with other ordinary bearings,truck bearings are difficult to accurately calculate due to complex structure,large bearing capacity,difficulty in accurately calculating the characteristic frequency of faults,and huge noise in the train operating environment,which brings great difficulties to safety monitoring and fault diagnosis.In this paper,taking the railway wagon 353130 B double row tapered roller bearing as the research object,the Variational Mode Decomposition(VMD)method is selected to realize the noise reduction processing,and according to the parameter sensitivity characteristics of VMD,the intelligent optimization algorithm is used to optimize the number of modes and penalty factors of VMD,construct a data set,and combine with a one-dimensional residual neural network model for fault diagnosis.The main research contents of the paper are as follows:(1)Since the number of modes and penalty factors of VMD decomposition have a great influence on the decomposition effect,it is proposed to use the envelope entropy as the fitness function,and the beluga optimization algorithm is used to find the optimal parameters.The optimization effect is verified by bearing fault simulation signal,compared with Empirical Mode Decomposition(EMD),and finally the effective component is selected for reconstruction by taking the relevant steepness as an index,and the superiority of VMD in noise reduction of rolling bearing fault signal is proved by envelope spectrum analysis.(2)Aiming at the complex working conditions of rolling bearings,the low diagnostic accuracy of traditional fault diagnosis methods,excessive reliance on expert experience,and the loss of some characteristic information when the bearing one-dimensional signal is converted into two-dimensional data,a rolling bearing fault diagnosis method based on one-dimensional residual neural network is proposed.This method takes the reconstruction signal of Beluga Optimized VMD as input,and adds a jump residual connection,which enhances the learning efficiency of the residual block on the feature information and can effectively extract the feature information.(3)Taking the railway wagon 353130 B bearing as the object,the railway bearing dynamic performance test bench of Dalian Jiaotong University was used to collect the fault conditions under various working conditions and the vibration signal under normal working conditions,and the effectiveness and generalization of the proposed fault diagnosis method were verified.The experimental results show that the fault recognition rate of the dataset after optimizing VMD noise reduction using beluga whale reaches 99.4%,which is significantly better than the original dataset of 91.67%.Compared with the classical convolutional neural network model,the proposed model has great advantages in both accuracy and convergence speed.Compared with the data sets under different speed conditions,the fault identification rate reached more than 98.5%,which further shows that the model can be used for the fault recognition of railway freight car bearings and has good generalization ability.