Electronic system fault diagnosis with optimized multi-kernel SVM by improved CPSO

Electronic systems' safety operation has become a key issue to complex and high reliability systems. Now more emphasis has been laid on the accuracy of electronic system fault diagnosis. Based on the characteristics of the electronic system fault diagnosis, we design a multi-classification SVMS model to attain better fault diagnosis accuracy, which utilizes multi-kernel function consisting of several basis kernel functions to enhance the interpretability of the classification model. In order to optimize the performance of multi-classification SVMS with multi-kernel, we improve the Chaos Particles swarm Optimization (CPSO) algorithm to achieve the optimum parameters of SVMS and the multi-kernel function. For the improved CPSO algorithm, a modified Tent Map chaotic sequence is used to strengthen the search diversity, and an effective method is embedded to the stander PSO algorithm which can ensure to avoid premature stagnation and obtain the global optimization values. The fault diagnosis simulation results of an electronic system show the proposed optimization scheme is a feasible and effective method and it can signifcantly improve the fault diagnosis accuracy of the multi-kernel SVM.