基于灵敏度分析及Volterra级数的非线性模拟电路故障诊断  

Fault diagnosis of nonlinear analog circuits based on sensitivity analysis and Volterra series

作　　者：韩海涛 马红光[1] 谭力宁[1] 张家良[2] 

机构地区：[1]第二炮兵工程大学101教研室,陕西西安710025 [2]西安交通大学机械制造系统工程国家重点实验室,陕西西安710049

出　　处：《电路与系统学报》2013年第1期17-22,共6页Journal of Circuits and Systems

基　　金：国家自然科学基金(61174207;61074072)

摘　　要：为了有效解决非线性模拟电路故障诊断问题,提出了一种基于灵敏度分析及Volterra级数理论的故障诊断方法。首先采用灵敏度分析估计激励信号的有效频带范围,根据Volterra级数确定激励信号的频率成分,以使故障可测度最大化;其次基于多音激励下Volterra核的频域输出特性进行故障特征提取,并采用主元分析(principal componentanalysis,PCA)对故障特征进行维数压缩;最后构造多值分类支持向量机(support vector machine,SVM)网络进行故障模式判别。理论分析及仿真结果表明,与以往方法比较,该方法可以显著提高非线性模拟电路故障识别率,验证了该方法的有效性。To effectively solve the fault diagnosis of nonlinear analog circuits,a novel fault diagnosis method is proposed based on sensitivity analysis and Volterra series.Firstly,the effective frequency band range of the input stimuli signal is estimated by sensitivity analysis,then the frequency components of stimulus signal are determined in terms of Volterra series,which make fault observability obvious.Secondly,fault signatures are extracted by virtue of the frequency output properties of Volterra kernels stimulated by the multitone signal,then they are compressed with principal component analysis（PCA）.Finally,the multiclass support vector machine（SVM） network is constructed for classifying fault types.The theoretical analysis and simulation results indicate that,compared with former method,the proposed method can observably increase fault recognition rate of nonlinear analog circuits,which validates effectiveness of the proposed method.

关 键 词：灵敏度分析 VOLTERRA级数 主元分析 支持向量机 多音信号 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]