Research On Analog Circuit Fault Diagnosis Technology Based On Support Vector Machine

The analog circuits are an important part of the electronic device,the reliability of circuit components directly can affect the operation of electronic devices in the entire system.However,the integration of the circuits is getting higher and higher,the circuit components in the analog circuits are nonlinear and tolerant,at the same time external factors can lead to fault diversity problems,making the analog circuit fault more and more prominent,so the research on analog circuit fault diagnosis has been particularly important.Support vector machine(SVM)is a learning algorithm with better generalization ability in the case of finite samples,which can effectively solve the problem of analog circuit fault diagnosis classification.Therefore,SVM is used to study the fault diagnosis of analog circuits.This paper mainly studies three aspects: method of fault feature extraction,optimization of kernel function,improved SVM algorithm.In order to solve the problem of fault feature extraction,four feature extraction methods are studied: principal component analysis,wavelet analysis,wavelet packet analysis and S transform.The principle of each method is analyzed theoretically,and the simulation is carried out by the example circuit,and the support vector machine model is used to classify and diagnose.According to the final fault diagnosis rate,the characteristics,application advantages and fault classification effect of each extraction method are analyzed.The penalty parameters in the SVM classifier and the kernel parameters of the radial basis function seriously affect the fault diagnosis performance,therefore,two parameters are optimized by double-chain quantum genetic algorithm.Two common datasets named Iris and Wine from University of California-Irvine(UCI)are used to test the advantages of the method,which is that the optimization performance is good,the classification precision is high,the classification rate is fast.The simulation experiment is carried out by two analog circuits,circuit 1 is a multi-fault situation and the use of wavelet method to extract the fault feature,the circuit 2 is a single soft fault and the use of S-transform time-frequency relationship to obtain eigenvalues.And compared with the traditionalgenetic algorithm,particle swarm optimization and quantum genetic algorithm,it is shown that the optimization method can obtain higher optimal fitness and shorter convergence time.Aiming at the problems of fault identification and low diagnosis accuracy of analog circuit,A fault diagnosis method based on K neighbor and support vector machine classifier is used.This method can effectively solve the problem of misclassification caused by the non-separable SVM and improve the fault diagnosis rate.Wavelet transform is used to extract the fault features of the output voltage signal,the grid search is adopted to optimize kernel function and penalty parameter.The simulation experiment is carried out by two analog circuits,the improved SVM is compared with the traditional SVM,the simulation results show the feasibility of the algorithm.