Fault Diagnosis of Analog Circuits Based on IH-PSO Optimized Support Vector Machine

Because of its excellent small sample learning abilities and simple network structure, support vector machine (SVM) is widely applied in various pattern recognition fields, e.g., face recognition, scene classification, fault diagnosis, etc. Due to the complexity and diversity of analog circuit faults, the diagnosis accuracy and stability of SVM classifier optimized by traditional particle swarm optimization (PSO) are unsatisfactory. Therefore, this paper proposes an improved hybrid particle swarm optimization (IH-PSO) algorithm to optimize SVM, which is applied in the fault diagnosis of analog circuits. Compared with the traditional PSO algorithm, the proposed IH-PSO mainly has three improvements, namely, the opposition-based learning population initialization, the nonlinear time-varying inertia weight, and the new position updating strategy with a spiral convergence mechanism. The performance of the proposed IH-PSO algorithm is verified by 12 commonly used benchmark functions and experimental results show that the proposed IH-PSO algorithm overcomes the deficiencies of the traditional PSO algorithm, such as slow convergence speed and trapping into local optimums. In addition, to further verify the performance of IH-PSO algorithm, the IH-PSO optimized SVM is applied to solve analog circuits fault diagnosis problems. Extensive experiments are carried out and results indicate that the proposed method has better performances both in diagnosis accuracy and stability compared with that of the traditional method.