基于同步优化的支持向量机模拟电路故障诊断方法研究  

Research on Method of SVM Analog Circuit Fault Diagnosis Based on Simultaneous Optimization

作　　者：申宇皓[1] 孟晨[1] 张磊[2] 傅振华[2] 石纯[3] 

机构地区：[1]军械工程学院导弹工程系,河北石家庄050003 [2]军械工程学院科研部,河北石家庄050003 [3]军械工程学院政治部,河北石家庄050003

出　　处：《太原理工大学学报》2010年第4期420-424,444,共6页Journal of Taiyuan University of Technology

摘　　要：支持向量机模拟电路故障诊断涉及到特征提取、特征选择和支持向量机的参数优化等问题,它们都对诊断结果有直接的影响。针对这一问题,提出了一种基于改进的离散粒子群算法的同步优化方法。该算法采用非线性惯性权重和遗传操作相结合的方法,提高了粒子群前期迭代的探索能力和后期迭代的开发能力,同时降低了粒子群陷入局部最优的风险。通过模拟电路的仿真实验,验证了同步优化方法和改进的离散粒子群算法的有效性。Analog circuit fault diagnosis based on SVM includes feature extraction,feature selection and SVM parameter optimization,all of which have direct influence on diagnosis result.A simultaneous optimization method is introduced,which is based on Improved Discrete Particle Swarm Optimization.Improved Discrete Particle Swarm Optimization combines nonlinear inertia weight with genetic operation,to enhance exploration in early stage and exploitation in late stage,and reduce the risk of particle swarm into local optimum.A fault simulation experiment of some analog circuit shows that this method is effective.

关 键 词：小波包变换 支持向量机 模拟电路故障诊断 离散粒子群 

分 类 号：TN707[电子电信—电路与系统]