| With the development of China's railway industry,the high-speed railway has become China's business card and is going to the world.The train has made great progress in speed and stability.Of course,while constantly pursuing speed improvement,we should also consider the issue of security.The safety of locomotive and rolling stock transportation is affected by many aspects.The axle is an important part of locomotive and rolling stock.Most of the axle failure is caused by fatigue cracks.Fatigue cracks can lead to axle damage or fatigue fracture,which leads to accidents.In order to solve the problem of fatigue crack detection of train axles,the acoustic emission technology is used in this paper.A method for analyzing experimental data of Fatigue Crack Acoustic emission signals of DTCWT-DBN is proposed.The key of the research is the feature extraction and recognition of acoustic emission signals.Wavelet transform has been successfully applied in signal extraction.The characteristics of dual-tree complex wavelet(DTCWT),discrete wavelet(DWT),empirical mode decomposition(EDM)and ensemble empirical mode decomposition(EEMD)are compared.Finally,the dual-tree complex wavelet is used to preprocess the acoustic emission signal to reflect the fault feature information.At the same time,the deep belief network is used to classify and identify the possible fault states of train axles,and the principal component analysis is used to classify the possible fault states.Data visualization of hidden layers of deep belief network is carried out,and the above feature extraction method and deep belief network are combined to apply to the fault diagnosis of train axle fatigue crack.Through feature extraction and recognition results of AE signal of train axle,it can be seen that the fault recognition method based on dual-tree complex wavelet and depth belief network has a higher recognition rate for the type of axle signal,and has a higher recognition rate compared with BP and SVM. |
PDF Full Text Request