基于改进支持向量机的模拟电路故障诊断研究

针对支持向量机（Support Vector Machine）及小波分解用于模拟电路故障诊断时,一对一算法具有操作简单、诊断精度高、所需确定参数少,小波分解能表现电路响应特征但最优小波基选取目前缺乏有效方法的特点,提出利用混合粒子群算法（Hybrid Particle Swarm Optimization,HPSO）对小波基及一对一支持向量机的参数进行联合优化。将该方法应用于模拟滤波器的仿真电路实验,结果表明：利用该方法很容易求出全局最优解,能实现对最优小波基选取及支持向量机参数进行联合优化,避免了参数选择的盲目性,提高了模型的诊断精度。

Research on Circuit Fault Diagnosis Based on Improved SVM

Since because there is not a mature method for choosing the parameters of SVM and wavelet function,a hybrid particle swarm optimization（HPSO） method is introduced to choose wavelet function and parameters of one versue one support vector machines（OVO-SVM） at the same time.The simulation results show that the fault diagnosis method has good diagnosis effect and feasibility in analyzing the fault response of analog circuits and can locate the faults in analog circuits correctly.