Experimental Data Analysis Of Acoustic Emission Signal Of Axle Fatigue Crack Based On DBN-BiLSTM

The axle is the key component of the bogie structure of the rail vehicle.Under the condition of long-term external load,it is easy to produce fatigue cracks.With the further expansion of the cracks,it will eventually lead to the fracture of the axle,which will lead to failure.Therefore,monitoring of axle status in real time,identify the fatigue crack signal of the axle,find out the fault in time and predict the life of the axle,all of them are of great significance.In order to identify the acoustic emission signal of axle fatigue crack from noise signal and knocking signal,a method of identifying the acoustic emission signal of axle fatigue crack based on Teager energy operator and Deep Belief Network is proposed in this paper.The experimental results show that the TEO can retain the instantaneous energy of the signal,and the fatigue crack signal of the axle can be identified from the noise and knocking signal through the classification of DBN network model.Compared with the traditional classifier,DBN has better recognition effect on the acoustic emission signal of axle fatigue crack.According to the process of axle fatigue crack initiation,propagation and fracture,the signal of axle fatigue crack is processed by using SK index.In order to simplify the structure complexity of the network and save the training time of the network,the extracted feature data set is used to replace the original data as the input of the network,and the DBN network is used to identify the different stages under the interference of noise and knocking signal.In order to accurately identify the different stages of fatigue crack and further improve the identification accuracy of each stage,a method of identifying the acoustic emission signal of axle fatigue crack in different stages based on SK derived features and BiLSTM network is proposed in this paper.Through the comparative experimental analysis,the results show that the method of SK derived features combined with BiLSTM network can effectively identify the acoustic emission signals of axle fatigue cracks in different stages;the fault state of fatigue crack of axle can be expressed more accurately by adding SK derived feature;compared with other network models,BiLSTM network model is more stable.Finally,through the verification experiment,it is verified that the BiLSTM network model has certain applicability for the identification of different stages of axle fatigue cracks.