Experimental Data Analysis Of Acoustic Emission Signal For Fatigue Crack Of Axle Based On LMD-SDAE And SAE

In recent years,the country has given great support to the rail transit industry,which makes the railway locomotives in our country also have a considerable development.As the core component of bogie,the safety of train axle is always the top priority of train safety.It is of great significance to be able to detect axle faults effectively and timely.However,the traditional fault signal processing methods are usually weak for non-linear and non-stationary signal processing.And there are some deficiencies in dealing with a large number of data.So this paper proposes a fault signal classification and recognition method based on the combination of LMD and sdae.Three kinds of fault signals are collected through the acoustic emission experiment of fatigue crack of train axle,which are fatigue crack signal of axle,background noise signal and knock signal of axle.Firstly,LMD is used to decompose the collected three kinds of fault signals,and 14 kinds of eigenvalues are calculated by the first three pf components of energy proportion,then the eigenvalues are input into sdae to train the network model,and the trained network model is used to classify and identify the three kinds of fault signals of locomotive axle,and then the performance of the network model is tested by the test set data High classification accuracy.The generalization ability of the network model is verified by the validation set data.Compared with the deep confidence network(DBN)and the limit learning machine(ELM),the recognition rate of this method is increased by about 7%.Compared with single sdae method,it can greatly shorten the training and testing time.The original data of AE signal of fatigue crack of axle is used to set sliding window to group the data,and the grouped data is divided into training set and test set.The training set is used to train sparse automatic encoder(SAE)prediction network model,and the test set data is used to evaluate the performance of prediction network model.By comparing the prediction network model of SAE with that of ssae,the conclusion shows that the prediction model of SAE can predict the acoustic emission signal waveform of fatigue crack accurately..