基于非线性电路频域核估计和神经网络的故障诊断

提出一种基于非线性电路频域核分析和神经网络的故障诊断方法．主要研究非线性系统频谱的获取,非线性系统频谱特征的提取及基于非线性系统频谱特征的故障诊断．利用Volterra频域核估计辨识非线性系统，通过系统广义频率响应函数的估算提取电路特征，将其预处理后作为递归神经网络的输入样本，利用神经网络的分类功能对电路的工作模式作出故障决策、最后。给出故障诊断实例验证了该方法的有效性．

Fault diagnosis based on neural network and frequency-domain kernel estimation of nonlinear circuit

A fault diagnosis method based on frequency-domain kernel analysis in nonlinear circuit and neural network is presented. Three contents are researched which include the frequency spectrum acquisition of nonlinear system, the frequency spectrum feature extraction in nonlinear system, and the fault diagnosis based on frequency spectrum feature in nonlinear system. A nonlinear system is identified by applying yolterra frequency-domain kernel estimation. The generalized frequency response function is computed and estimated to extract circuit features which are pre-processed as input samples in internet recurrent net. Circuit work modes are classified by neural network. An example shows the effectiveness of the proposed method.