Surface Stress And Hardness Non-Destructive Measurement Of Ferro-magnetic Material Based On Barkhausen Effect

Magnetic Barkhausen noise(MBN)technology is a novel electromagnetic NDT method,which based on the theory of magnetization effected by stress distribution and microstructure.This technology is considered as an effective method for early damage and microstructure testing and characterization.But the traditional features ignore the details of magnetization,microstructure effect,and their complex relationships.Thus,MBN features are easy to be influenced by the excitation frequency,which makes the technology hard to characterize the microstructure accurately.In order to explore the problems above,the effects of excitation parameters on MBN features are studied firstly.Then a new feature based on domain wall dynamics is provided and discussed to reduce the influence of excitation parameters.After that,the research results are applied for surface hardness detection of carburized metal.A new method based on correlation analysis and principal component analysis(PCA)is proposed for feature selection.At last,Multiple Linear Regression(MLR)is applied for surface hardness modeling.The main contents and contributions of this thesis are listed as below:1)The theories of domain energy and domain wall(DW)dynamics effect on MBN are summarized.Based on the theory,the excitation parameters effects on MBN signals are studied.Considering MBN signals affected by the material microstructure obviously,the optimization of excitation parameters is proposed based on the sensitivity comparison of MBN features.This optimization is the research foundation for surface stress measurement and hardness evaluation using MBN technology on different material.2)The conventional feature extraction only in view of the MBN signal energy or the height/width of MBN signal profile.To obtain much information from different dimensionality,skewness of MBN signal profile is proposed as a new feature based on the theory of domain wall dynamics.The new feature characterizes applied stress and variation of material microstructure by the distribution of Barkhausen pulses which reflect the DW motion property during magnetization.Due to the MBN signal energy and time-domain distribution,Skewness is more sensitive than conventional features and independent of the excitation frequency.Based on the theory of DW energy and dynamics,the relations between applied stress and MBN signal skewness are discussed.The physical model of relaxation time of DW increases with the improvement of pinning density is proposed,which illuminates the non-linear relations between skewness and applied compressive stress.3)Considering single MBN feature can not evaluate surface hardness accurately,multivariate linear regression(MLR)modeling method based on MBN features selection is proposed to improve the accuracy of hardness evaluation.Principal component analysis combined with parameters correlation analysis is proposed as a new method for feature selection,which has higher efficiency and more global information than conventional methods.Based on this method,root-mean-square(RMS),mean and delay-time of MBN signal are selected for material surface hardness modeling using MLR method.During this process,the carburized material hardness is predicted accurately.In summary,studies in this thesis provide a novel feature extraction based on the theory of Barkhausen effect and DW motion.This new feature,skewness,can reduce the influence of excitation frequency and build up the relationship between the electromagnetic signals in macroscale and the magnetic microstructure of materials.On the other hand,this thesis provides an improved PCA method for feature selection and employs the selected features for building up material surface hardness MLR model.Then the stress distribution and material surface hardness are evaluated accurately by MBN technology.At last,the further work is summarized based on the conclusions of the studies in this thesis.