基于非线性频谱特征及核主元分析的模拟电路故障诊断方法

针对模拟电路基于非线性输出频域响应函数(NOFRF)模型进行故障特征提取时,具有维数多、数据量大的特点,提出了采用核主元分析(KPCA)和多类别支持向量机(MSVM)进行故障模式判别的新方法(KPCA-MSVM)。该方法首先采用KPCA对特征向量进行维数压缩、消除变量之间的非线性;其次构造MSVM分类器,在PSpice环境下通过蒙特卡罗仿真生成模拟电路在各种故障状态下的数据,对MSVM分类器进行训练,将训练好的MSVM分类器用于模拟电路的故障状态识别。通过对Sallen-Key带通滤波器模拟电路的故障诊断结果表明,该故障诊断方法对模拟电路参数型故障有很好的识别、定位能力并具有速度快和准确率高的特点。

Fault Diagnosis Method of Analog Circuits Based on Characteristics of the Nonlinear Frequency Spectrum and KPCA

For the characteristics that there existed much dimensions and big data volume in extracting fault signatures based on the model of nonlinear output frequency response function(NOFRF),a novel fault diagnosis method,which adopted kernel principal component analysis and multi-class support vector machine(KPCA-MSVM),is proposed to identify different fault states.Firstly,kernel principal component analysis(KPCA)is used to compress data dimension and eliminate nonlinearity among the variables.Secondly,multi-class support vector machine(MSVM)classifier is constructed,and the datum of all kinds of fault states,which were used to train MSVM classifier,are generated by Monte Carlo simulation with PSpice software.The trained MSVM classifier is used to identify different fault state.Via fault diagnosis for sallen-key band pass filter,the results indicate that KPCA-MSVM has good ability to identify and locate parametric faults for analog circuits,and has virtues of fast speed and high precision.