基于Volterra频域核的非线性模拟电路故障诊断  

Nonlinear Analog Circuit Fault Diagnosis Based on Volterra Frequency-Domain Kernel

作　　者：康辉[1] 黄耀群[1] 王金霞[2] 

机构地区：[1]黑龙江科技学院电气与信息工程学院,黑龙江哈尔滨150027 [2]哈尔滨技师学院电气技术系,黑龙江哈尔滨150030

出　　处：《测试技术学报》2011年第4期366-370,共5页Journal of Test and Measurement Technology

摘　　要：非线性模拟电路的故障特征的表征困难,另外基于神经网络的故障诊断存在收敛速度低及泛化能力差的问题.本文将非线性系统复杂的Volterra高阶核函数表示成一维数值解的形式,并以此来表征故障特征,同时借助小波变换与神经网络的优点,降低了输入特征矢量的维数,并且提高了泛化能力.实验结果表明:该方法建模简单,运算量小,具有一定的可行性和实用性.The characterization of the nonlinear analog circuit fault feature is difficult, The convergence rate of fault diagnosis based on neural network is low and there is the problem of poor generalization ability. We express the complex nonlinear system Volterra higher-order kernel function as the one- dimensional numerical solution to represent the fault feature. The wavelet transform and neural network not only reduce the dimen- sion of input feature vectors, but also enhance the generalization ability. Experimental results show that the method is simple, feasible and practical in operation.

关 键 词：非线性模拟电路 故障诊断 VOLTERRA核 小波变换 神经网络 

分 类 号：TN911.7[电子电信—通信与信息系统]