The Acoustic Emission Signal Feature Modeling And Analysis Of Fatigue Crack Of Gear Material

Fatigue crack of mental material appears under stress and strain in long-term,the damage constantly accumulates and eventually breaks,which is fatigue failure.Sudden fatigue break of parts of mechanical equipment could cause significant accidents and loss.20 CrMnTi has good comprehensive properties and it is a widely used gear material in our country.When the mental material appears crack,the energy releases quickly and transient elastic wave appears,which is acoustic emission phenomenon.The research studied the fatigue crack propagation of 20 CrMnTi test bar,used acoustic emission technology to gather the acoustic emission signals of rotating bending fatigue of test bar under no-load condition and load condition.In order to identify the acoustic emission signals of test bar under load condition,parameter analysis and time-frequency analysis were carried out,BP neural network and SOM neural network were used to classify and cluster the acoustic emission signals.This paper draws the following conclusion:(1)Energy count,ring count,effective value RMS and average signal level ASL of acoustic emission signals under load condition were higher.(2)Hilbert spectrum of the acoustic emission signals of no-load condition and load condition of two experiment conditions had no obvious differences.(3)13 parameters of the acoustic emission signals were taken as BP neural network inputs,the acoustic emission signals were divided two types by energy count threshold value 55 and 35,class A were noisy signal and class B were acoustic emission signals of test bar,which were taken as the expected outputs of BP neural network after training,400 acoustic emission signals were divided two types by BP neural network,accurate rates of recognition of two kinds of acoustic emission signals were both very high.(4)16 parameters of the acoustic emission signals were taken as SOM neural network inputs,the acoustic emission signals under load condition were clustered two types,energy counts of 7 signals of class A were all zero,only one energy count of 43 signals of class B was zero.The 50 mixed acoustic emission signals under two experiment conditions were clustered two types,effective value RMS of 24 signals of class A were in 0.08~0.09 and effective value RMS of 26 signals of class A were in 0.12~0.13.This paper has the following innovative points:(1)It was carried out time-frequency analysis of acoustic emission signals using HHT,the acoustic emission signal frequency-time joint distribution was got.(2)The acoustic emission signals were classified and clustered by artificial neural network,it was gained good effect,which laid the foundation of identifying valuable acoustic emission signals in the complex environment.