Research On Early Fault Diagnosis And State Recognition Method Of Urban Rail Train Rolling Bearings

As an important mode of public transportation,urban rail transit can solve the problems such as traffic congestion and air pollution.In recent years,urban rail transit has gradually become the lifeblood of the major cities of China.However,in the case of rail transit is increasingly favored by people,train safety issues are increasingly important.As the key components of the train,rolling bearings are also fault multiple components,whose running state directly affects the safe operation of the train.Therefore,real-time monitoring and analysis of rolling bearings and accurately grasping the working state of rolling bearings are of great significance for preventing accidents and ensuring reliable operation of trains.Given this,in this paper,the methods of the early fault diagnosis and state recognition of the rolling bearings are studied.The main contents are as follows:(1)The early fault diagnosis method of rolling bearings based on variational mode decomposition(VMD)is studied.Aiming at the problems of greatly influenced by noise and prone to modal aliasing of traditional empirical mode decomposition(EMD)when processing signal,this paper proposes VMD method to analyze the early fault signal of rolling bearings.The influence of the selection of key parameters in the VMD algorithm on the results is studied,and the chaotic particle swarm optimization(CPSO)is improved to make it suitable for VMD parameters optimization.Through analyzing and comparing the rolling bearings early fault simulation signal and the whole life fatigue accelerated test data,it is proved that the proposed method can effectively identify the early weak faults of rolling bearings,and has more advantages than traditional methods.(2)The feature extraction technology of vibration signal of rolling bearingss is studied.In this paper,the time domain characteristic parameters of rolling bearings vibration signal are extracted first.Then the double-tree complex wavelet packet transform(DT-CWPT)is used to decompose the signal,the multiscale permutation entropy(MPE)of the coefficient of node reconstruction is obtained.In order to avoid the adverse effects of feature redundancy on the recognition results,by random forest algorithm(RF)is used to do feature selection and characteristic parameters with higher importance are selected as the final input set of the modal identification algorithm.(3)The mode recognition method based on KELM-AdaBoost of rolling bearing is studied.Based on the basic extreme learning machine(ELM),this paper focus on the kernel-based extreme learning machine(KELM)algorithm.CPSO algorithm and cross validation method are combined to optimize the parameters of KELM.In view of the poor effect of single KELM classification,an integrated learning algorithm based on KELM-AdaBoost is proposed to improve the performance of the model.By analyzed on the rolling bearing vibration signals,the experiment results showed that the method can effectively identify the rolling bearings with different fault types and different damage degree.