Recognition And Analysis Of Acoustic Emission Signal Of Axle Fatigue Crack Based On PSO-DBN

As a major railway transportation country,China’s railway transportation occupies an important position in the whole transportation industry.Ensuring the safety of railway transportation is not only an urgent task for national development,but also the ardent expectation of the people all over the country.In recent years,there have been many serious traffic safety accidents in China’s railway,especially the major fatal accidents of passenger trains,which have caused great losses to China’s economy and people’s life and property.Axle is one of the important parts of train running parts.Due to its complex service environment,the axle is very easy to produce fatigue cracks.After the crack initiation and propagation to a certain stage,the axle will break,resulting in extremely serious consequences.From this,it can be seen that the real-time monitoring of the running state of the axle is extremely important.This paper focuses on the optimization of the number of hidden layer nodes of deep belief network(DBN)by particle swarm optimization(PSO),and uses the optimized DBN to classify and identify the axle acoustic emission signals.At the same time,the axle acoustic emission signal recognition system is preliminarily designed and developed.This paper firstly introduces the structure and basic principles of DBN and PSO.On this basis,the network model of PSO optimizing the number of hidden layer nodes of DBN is established,and a new DBN network is constructed with the number of hidden layer nodes obtained after PSO optimization,which is used to classify and identify the axle acoustic emission signals.PSO-DBN,extreme learning machine,one-dimensional convolutional neural networks and original DBN are used to classify and identify axle acoustic emission signals under the same working condition.The classification accuracy of the four network models can be compared to verify the advantages of PSO-DBN in classifying and identifying axle acoustic emission signals.Then,the PSO-DBN model is used to classify and identify the axle acoustic emission signals under different working conditions.The research results show that the PSO-DBN model can well classify and identify the axle acoustic emission signals under different working conditions.Finally,this paper makes a preliminary design and development of the page design of the axle acoustic emission signal recognition system and the data processing part based on PSO-DBN,and preliminarily realizes the functions of registration and login,data acquisition,data storage,data processing and realtime monitoring.