Analogue Circuit Fault Diagnosis Based On SVM Optimized By Swarm Intelligence Algorithm

With the increasing complexity and integration of analog circuits in modern electronic devices,the reliability and maintainability of the electronic devices are critical.However,analog circuits have problems such as less measurable nodes,high redundancy of circuit information,tolerance,nonlinearity and diversity of faults,which makes analog circuit fault diagnosis develop slowly.In this paper,support vector machine(SVM)is applied to analog circuit fault diagnosis,but its penalty parameter C and kernel parameter g have different influence on classification performance.To solve this problem,this paper propose an improved particle swarm optimization algorithm and an improved gravity search algorithm to optimize the parameters of SVM respectively,and simulate and verify the linear and nonlinear analog circuits.The main contents of this paper are as follows:First,the fault diagnosis method of analog circuit and the swarm intelligence algorithm used to optimize the SVM parameters are studied.By analyzing the characteristics of analog circuit faults,a fault diagnosis model is built based on SVM.Wavelet packet decomposition is used to extract feature,and swarm intelligence algorithm is used to optimize two parameters of SVM to get the best fault classification effect.Secondly,particle swarm algorithm which is easily fallen into local convergence is improved by introducing a new dynamic inertia weight,global neighborhood search,shrinkage factor and mutation operator of genetic algorithm,and this paper proposes a modified classifier that uses the improved particle swarm optimization(IPSO)tooptimize the parameter of SVM(IPSO-SVM).The simulation and experiment of fault diagnosis are carried out by linear and nonlinear analog circuits.The results show that the IPSO-SVM classifier has better classification performance than GS-SVM,AFSA-SVM,GA-SVM and PSO-SVM classifier,and has strong global convergence ability and faster convergence speed.Finally,in order to further improve the efficiency of analog circuit fault diagnosis,shorten the time of classification,this paper proposes a modified classifier that uses the improved gravitational search algorithm(IGSA)to optimize the parameters of SVM(IGSA-SVM)by introducing the inertia weight and global memory in particle swarm algorithm,time-varying gravitational search strategy and boundary mutation strategy.Three UCI data and two analog circuits are used to verify the classification results of IGSA-SVM classifier,the results show that the IGSA-SVM classifier can effectively prevent local convergence and shorten the classification time of fault diagnosis.