Research On Fault Diagnosis Method Of High-speed Railway Train Wheel Set Bearing Of Based On Deep Learning

In recent years,with the continuous increase of high-speed railway operating mileage and the number of trains,high-speed railway operation safety has attracted more and more attention.As one of the key power transmission components of the train,the wheel set bearing is used frequently and intensively.Whether its working state is normal or not directly affects the operation safety of the high-speed train.Therefore,it is of great significance to study reliable fault diagnosis methods for wheel set bearings to ensure the safety of high-speed railway operation.Taking the wheel set bearing of high-speed railway train as the research object,this paper studies the problems of noisy data,uncertain operating conditions and few fault samples of the collected data in the train operating environment,and puts forward two bearing fault diagnosis models and a cross domain bearing fault diagnosis model of transfer learning.(1)In order to solve the problem that the deep learning model can not extract the characteristics of high-speed train signals because of complex working conditions and noise in vibration signals,a bearing fault diagnosis model based on improved one-dimensional convolutional neural network is proposed by using multi-scale receptive field and compression and excitation module.In the network,multi-scale receptive fields are used to extract multi-dimensional features.The channel attention mechanism of Se module learns the feature channel weights,and recalibrates the weights to higher weights of useful features,so as to improve the accuracy of the final classification results.Using Case Western Reserve University bearing vibration signal data set and high-speed railway wheel set bearing vibration signal data set,it is verified that the proposed model is effective in feature extraction and fault classification.(2)In order to solve the problem that the accuracy of bearing fault diagnosis is low under the variable operating conditions of high-speed train,the residual block and channel mixed washing operation are introduced.Based on the model in the previous chapter,the network is further improved,and an improved one-dimensional convolutional neural network bearing fault diagnosis model based on feature fusion is proposed.The residual block and channel shuffling operation improve the ability of the network to extract features,and make the network stable enough to carry out fault diagnosis tasks under variable operating conditions.Experiments are carried out with the data sets of variable conditions constructed by the two data sets,and the proposed model is verified.For the data sets of variable conditions,the fault diagnosis and classification task can be effectively carried out.(3)Aiming at the problem that the train bearing is in normal state for a long time,the collected data fault samples are few,and the deep learning model is difficult to train,a cross domain bearing fault diagnosis model based on correlation alignment method is proposed.The multi-core maximum mean difference and correlation alignment method are used as the domain loss function.In the previous chapter,the domain adaptation layer is added to the model network to build a cross domain bearing fault diagnosis model.Using the two data sets,the migration experiments of different working conditions between the same data set and the migration experiments of environmental data from the laboratory of Western Reserve University to the vibration signal data of wheel set bearings of high-speed railway are carried out respectively.The results show that the proposed model has good migration performance and can solve the problem that it is difficult to train the model due to the lack of fault samples to a certain extent.