Acoustic Emission Signal Identification And Analysis Of Axle Fatigue Crack Based On Improved One-dimensional CNN

As a transportation mode with high economy and large transportation volume,railway plays a pivotal role in the national transportation system and is also a national infrastructure for people’s livelihood.Train axle as one of the important part of railway vehicle bogie,it not only bears the train car body in the process of operation for continuous pressure,but also bears the turning operation under variable pressure,at the same time because of the influence of the operating environment and weather,axle makes easy to produce fatigue crack,without processing,the crack will gradually expand,Finally lead to the occurrence of broken shaft,causing great harm to people’s life and property.Therefore,it is urgent to explore a stable and efficient train axle detection method.In order to solve this problem,in this paper,by using acoustic emission technology combined with deep learning network,with one dimensional convolution based on neural network and the residual network presents a neural network based on improved onedimensional convolution axle fatigue crack acoustic emission signal identification method,the network model is mainly composed of large size of one-dimensional convolution pooling module and residual module,These two modules give consideration to the study of low frequency features and deep features of data.The experimental data in this paper is a data set of acoustic emission signals of axle fatigue crack mixed with background noise signals and knock signals.The experimental results show that the improved one-dimensional convolutional neural network method has a good recognition effect on the acoustic emission signals of axle fatigue crack.In order to compare the rationality and effectiveness of the method proposed in this paper,one-dimensional convolutional neural network,VGG16 network and LSTM network are selected for comparative experiments.The results show that the improved onedimensional convolutional neural network has better network performance.In the axle fatigue crack acoustic emission signal recognition after analysis,this paper,by using the Python language developed network identification module and real-time monitoring module,the train axle acoustic emission signal acquisition after operation in the process of the recognition and axle of the real-time monitoring function,has realized the deep learning network model in the practical engineering application in the field of fault diagnosis.