Circuit Fault Diagnosis Algorithm Based On Hybrid Particle Swarm Optimization Algorithm Of Support Vector Machine

Circuit fault diagnosis is one of research fields of pattern classification.Support vector machines(SVMs)are widely used in information classification due to their good ability of generalization.However,the accuracy of classification depends largely on the kernel parameters and penalty parameters of SVMs.In terms of the optimal combination of kernel parameter and penalty parameter in SVM,this thesis proposes a new hybrid particle swarm simulated annealing algorithm to optimize and support the structural parameters of SVM,so as to attain better fault diagnosis accuracy.The main research content of this article is as follows:(1)Monte Carlo analysis is performed on the selected circuits of different kind to obtain the information of original fault.Wavelet packet decomposition is used to extract energy information of waveforms in different frequency bands so as to construct fault eigenvector.(2)For the reason that the kernel parameters and penalty parameters of SVM largely affect its ability of generalization,this paper optimizes the structural parameters of SVM via the intelligent optimization algorithm.However,the traditional particle swarm algorithm is easily running into the local optimization,so this paper proposes a new hybrid particle swarm simulated annealing algorithm.The algorithm uses the basic particle swarm algorithm as the framework,and combines the simulated annealing algorithm with particle adaptive mutation to improve the algorithm's global optimization ability and convergence ability.Therefore,simulated annealing algorithm mechanisms and adaptive particle variation are introduced into the basic particle swarm algorithm.Hybrid particle swarm simulated annealing algorithm uses PSO optimization algorithm with compression factor to ensure the convergence of the algorithm,and adopts roulette gambling transmission strategy and particle adaptive mutation to avoid local optimum,and validates the convergence of hybrid particle swarm simulated annealing algorithm through typical test functions.It is made to verify the superiority and effectiveness of the proposed algorithm to comparison with the performance of improved particle swarm optimization algorithm and other typical algorithm optimization.(3)The kernel parameters and penalty parameters of the hybrid particle swarm simulated annealing algorithm to optimize SVM are applied to different kinds of circuit fault diagnosis.Typical analog circuits,nonlinear analog circuits and power electronic circuits are selected to conduct Monte Carlo analysis to collect original fault information.Using wavelet analysis to extract the energy of different frequency bands to constitute the fault feature vector,and divide them into training set and test set.The training set is used to train the SVM.The test set is used to verify the condition of fault classification.By comparing with other diagnostic algorithms,we intend to verify that the proposed algorithm has good diagnostic accuracy in analog circuit fault diagnosis,nonlinear analog circuit fault diagnosis and power electronic circuit fault diagnosis.Simulation and experiment results indicate that the proposed method indeed improves the accuracy of fault diagnosis.