基于神经网络的模拟电路快速故障诊断集群撕裂方法  

Neural Network-based Cluster Tear Approach for Fast Fault Diagnosis of Analogue Circuits

作　　者：郭荣斌[1] 王绪飞[1] 李中群[1] 

机构地区：[1]中国电子科技集团公司第41研究所,山东青岛266555

出　　处：《计算机测量与控制》2014年第4期1046-1049,1052,共5页Computer Measurement &Control

基　　金：国防科技计划项目(C1120110004;9140A27020211DZ5102)

摘　　要：提出一种基于神经网络的大规模模拟电路故障诊断集群撕裂新方法;将大规模模拟电路按照拓扑特性和集群撕裂准则进行撕裂,得到低维度的故障特征向量;利用小波神经网络实现故障特征向量的快速分类并得出电路诊断结果;通过对视频放大电路诊断事例,将电路划分12个子模块,进行两次撕裂与特征抽取,小波神经网络经过80次迭代即趋于稳定;结果表明,对于所测试例诊断正确率达100%,所提方法与传统互校验和交叉撕裂搜索法比较,具有测前工作量小,诊断计算量少,对多故障检测能力和工程实践性强等特点。A new approach for large--scale analogue circuits fault diagnosis using Cluster tear is presented. Low dimensional fault eigenvectors can be acquired by decomposing the large--scale analogue circuit in accordance with its topological property and Cluster tear criteria. To achieve the goal of fast classification of eigenveetors and obtaining diagnostic results rapidly, neural network which has highly parallel classify capability is selected as classifier and wavelet function which has fast convergence property is selected as hidden layer＇ s transfer excitation function. For example, by u- sing the presented approach for fault diagnosis of vadio amplifier circuit, through two times to tear the circuit into twelve sub--circuits, after feature extraction and 80 training of wavelet neural network, the results is in accord with the reality. This approach has smaller amount of work before test, fewer diagnostic times, less computation, better capability of diagnosing multiple faults and stronger engineering practicality than the existing multiple --test--condition （MTC） method and intersection tearing method.

关 键 词：大规模模拟电路 故障诊断 集群撕裂法 小波神经网络 

分 类 号：TP306[自动化与计算机技术—计算机系统结构]