Research On Soft Fault Diagnosis Of Analog Circuits Based On Multi-scale PCA

Modern electronic equipment has been developing rapidly,the complexity and integration of analog circuits in electronic power systems are constantly increasing.There are higher requirements for the reliability of modern large-scale equipment and system operation,which requires good maintainability of the analog circuits.In the sake of the cost and manpower,when the analog circuit break down,it is hoped that the faults generated in the circuit can be distinguished and accurately classified without complex calculations and incomplete grasp of circuit knowledge.Due to the lack of testable nodes in analog circuits,the inevitable tolerance of components and the non-linearity of most components,which lead to the diversification of analog circuit faults,the research on the soft fault diagnosis of analog circuits has never stopped.Aiming at the problem of soft fault diagnosis of analog circuits,this thesis mainly discusses the construction of accurate soft fault diagnosis model of analog circuits.Based on the framework of multi-scale principal component analysis,combined with the optimized support vector machine,the fault identification and fault classification of analog circuits are carried out correctly.In view of the shortage of current fault feature extraction methods,the multi-scale principal component analysis(MSPCA)method and its improvement are emphatically studied.The information of the high frequency part is added to its structure and the subband energy is introduced as the weight.The improved multi-scale principal component analysis method is used to reduce the dimension of the original signal data,and the fault feature information is extracted from multiple scales,which is used to improve the accuracy of analog circuit soft fault diagnosis.Then support vector machine(SVM)used for fault classification are analyzed in theory,two parameters that affect the classification effect of SVM are studied,and the method of using the swarm intelligence optimization algorithm to optimize and improve SVM is proposed.This thesis focuses on the analysis of firefly algorithm and its improvement,using the improved firefly algorithm to verify the classification effect on three common datasets of different sample sizes and dimensions.It is proved that the improved firefly algorithm is greatly improved compared with other algorithms from two aspects of classification accuracy and classification duration,which verifies its superiority in classification effect.Based on the method proposed in this thesis,the Sallen-Key bandpass filter circuit and IGBT driver circuit are built by using PSpice and Matlab to carry out experimental simulation and verification.By combining MSPCA with improved firefly algorithm optimized SVM for fault diagnosis,the fault diagnosis rate of both circuits reaches 100%,which is greatly improved compared with wavelet analysis and PCA.In order to verify the accuracy of the proposed method,several experiments were carried out on the two circuits respectively,and the average accuracy was close to 100%,which verified the effectiveness of the proposed method.