差分杂草算法优化SVM的模拟电路故障诊断  

Analog circuits fault diagnosis based on differential evolution invasive weed optimization support vector machine

作　　者：刘自鹏 李志华[1] 陈欣[1] LIU Zi-peng;LI Zhi-hua;CHEN Xin(School of Energy and Electrical Engineering,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学能源与电气学院,南京211100

出　　处：《信息技术》2017年第7期134-138,共5页Information Technology

摘　　要：针对非线性模拟电路软故障检测和定位难题,提出一种差分杂草算法(DEIWO)优化支持向量机(SVM)的故障诊断新策略。首先利用递推最小二乘算法对电路Volterra级数时域核进行辨识提取故障特征,然后用差分杂草算法(DEIWO)优化支持向量机参数建立故障诊断模型,后对故障进行分类识别,完成故障诊断。仿真结果表明,该方法具有较高的准确率。In order to solve the problem of soft fault detection and location of nonlinear analog circuits,a new fault diagnosis strategy based on support vector machine optimized by differential evolution invasive weed optimization algorithm(DEIWO)is proposed.Firstly,the kernel of volterra series are identified by recursive least squares method to extract the fault features.Then,differential evolution invasive weed optimization algorithm is used to optimize the parameters of support vector machine,and then it classifies the fault.The simulation results show that the proposed method is of high accuracy.

关 键 词：非线性模拟电路 Volteiia级数 差分杂草算法 支持向量机 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]