Study On Signal Processing And Defect Recognition Methods Of Pulsed Eddy Current Testing For Multi-layer Metal Lap Joint Structures

Multi-layer metal lap joint structures are widely used in aerospace,nuclear power and industrial production.Timely and reliable detection of possible defects in these structures is essential to eliminate potential safety hazards and guarantee the safety and stable operation of the system.This thesis mainly studies the pulse eddy current detection signal processing and defect recognition methods for multi-layer metal lap joint structures under the support of the National Science and Technology Major Project,the National Natural Science Foundation Project and actual engineering application requirements.Based on the principle of pulsed eddy current testing technology,the characteristics of pulsed eddy current response signal of multi-layer metal lap joint structures are analyzed by numerical modeling technique.The defect detection method,defect parameter identification method and scanning imaging technology of multi-layer metal lap joint structures are emphatically discussed.The pulsed eddy current detection system is integrated and improved.The main research work and innovations of the thesis are as follows:(1)The basic principle of pulsed eddy current testing technology is expounded and the defect models of multi-layer metal lap joint structures are established by application of finite element numerical modeling method.By calculation and analysis of pulse eddy current response signal and variation of magnetic field distribution,crack and corrosion response characteristics of multi-layer metal lap joint structures are detailed analyzed.The influence of excitation parameters on detection sensitivity,the influence of defects on current,magnetic induction intensity distribution and response curve,and the influence of the number of structural layers on the pulse eddy response signal are focused,which provides theoretical guidance to the pulsed eddy current inversion analysis.(2)On the basis of the existing analysis of pulsed eddy current detection signals,the issue of defect detection in multi-layer metal lap joint structures is studied and a method for defect discriminating by random forest algorithm based on combined features is proposed.Aiming at the problem that the change of time domain features under small cracks is weak and small cracks are difficult to be detected effectively only by time domain feature points,multi-dimensional features such as time domain peak point,frequency domain FFT amplitude,and wavelet packet energy in different frequency bands are carried out.Through feature combination and feature selection according to Gini coefficient,random forest algorithm is applied to realize the defect detection,which improves the detection rate in the small crack and lays the foundation for the subsequent qualitative and quantitative evaluation of defects.The applicability of the extracted features under different layers of metal lap joints is discussed.(3)On the basis of defect detection,the issue of defect parameter identification in multi-layer metal lap joint structures is studied and a method for feature extraction based on projection pursuit algorithm is proposed.Aiming at the problem that the nonlinear complex relationship between defect parameters and features and the traditional principle component analysis(PCA)dimension reduction method is not clear about the physical meaning,a feature extraction method based on AID index for projection pursuit algorithm is proposed to characterize defect parameters.It can preserve the inherent law of data while reducing data.The supported vector machine(SVM)and the support vector regression(SVR)method is applied to realize the position discrimination of defect and quantitative evaluation of defect's size,which improves the accuracy of qualitative and quantitative evaluation of small crack.The applicability of the projection pursuit method under different layers of metal lap joints is discussed and analyzed.(4)Around the issues of defect localization and imaging of multi-layer metal lap joint structures,the pulsed eddy current scanning imaging detection technology is studied.According to the characteristics of the continuous change in pulse eddy current response signal at the edge of the defect and the problem of weak change of edge information,the Canny operator is introduced to extract the defect edge of the scanned image.The localization in the horizontal direction of the defect is achieved by extracting the defect's edge.Aiming at the problem of the deviation between the extracted edges and the actual edges,an edge geometry correction method is proposed to improve the accuracy of edge extraction.Combined with the projection pursuit method in the point 3,the localization in the vertical orientation of the defect is achieved.After that,an automatic defect assessment method is proposed by combining the defect detection method and defect parameter identification method in points 2 and 3.(5)Improvement and optimization work is carried out based on the pulsed eddy current detection system in the laboratory.The signal acquisition and processing unit of the system is optimized design,which improves the scalability and stability of the system.The comprehensive experiment is carried out to verify improved pulsed eddy current testing systems and improved signal processing methods.The research work of the thesis improves the detection capability of pulsed eddy current testing technology and provides technical support for non-destructive testing and evaluation of important equipment.