Research On The Feature Extraction Of Acoustic Emission Signal Of Axle Based On Various Improved Wavelet Entropy

With the continuous progress of the times,science and technology develops day by day,more and more people choose rail transportation as a travel tool.At the same time,the condition monitoring and fault diagnosis of railway are becoming the focus of people's attention.The train axle is an important part to transmit power,and its fault status is a serious threat to the safety of the train operation.In view of this,the acoustic emission of train axle rotation signal contains lots information,use these information reasonably and effectively,is important to guarantee the safe and stable operation of the train,study on this has theoretical significance and engineering application value.This paper introduces the related knowledge of the acoustic emission detection technology?fault diagnosis technology?wavelet entropy,and the current research status at home and abroad firstly,aiming at the deficiencies of Shannon wavelet energy entropy and put forward the improve methods.Take the simulation experiment of axle rotation for a certain type of axle,signals are collected to provide the original data support for the research of signal processing methods in the following chapters.The first method is the Tsallis wavelet energy entropy.Firstly,discusse the extensive ductility of Shannon wavelet energy entropy,and proved that it is very difficult to analyze the signal with non extensive property.Then,the Tsallis wavelet energy entropy is proposed,by calculating the energy entropy of the normal signal and the fault signal,compare the differences,extract the feature of fault signal.Another methed in this paper is the wavelet packet singular entropy.Aiming at the weakness that the wavelet transform can not be used to decompose the high frequency signal constantly,the singular entropy method based on wavelet packet transform is proposed to extract the feature of the signal.Firstly,take the wavelet packet decomposition to the signal.Then the transformation of the reconstruction is processed by singular value decomposition theory.Finally,the entropy of singular eigenvalue is calculated.Analysis results show,signal components become complex due to crack failure,the entropy becomes larger.A large number of simulation experiments are carried out and the measured data are obtained,the results show the effectiveness of the proposed feature extraction methods and models,explore a new way to find the fault of axle by means of signal data analysis.