数字电路故障模块诊断的神经网络方法研究  

A Neural Network Approach to Block Level Circuits Fault Diagnosis

作　　者：潘中良[1] 熊银根[2] 

机构地区：[1]华南师范大学物理系,广东广州510631 [2]中山大学无线电电子学系,广东广州510275

出　　处：《中山大学学报（自然科学版）》2001年第4期51-54,共4页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：国家自然科学基金资助项目(60006002);广东省自然科学基金资助项目(001172)

摘　　要：对数字电路系统的测试，提出一种基于神经网络的电路故障模块诊断方法，对方法的诊断系统模型、故障特征的获取、网络结构的确定与调整、诊断的流程等进行了详细论述．由于传统的电路故障诊断方法是通过对故障字典的搜索来确定发生的故障，这对庞大的故障字典来说既复杂又费时。A technique for fault diagnosis of digital circuits and systems at block level is proposed. The Fourier transform methods are employed to obtain the features of logic blocks, the structure of neural networks which accord with the circuit under test is presented. The neural network gradually acquires and refines the knowledge of test samples via genetic evolution of the topology and weight distribution of network. The fault features can then be used as inputs to the trained networks, and the trained network is used to yield diagnosis results. The proposed technique enhances the efficiency of the conventional fault dictionary methods.

关 键 词：数字电路 故障诊断 神经网络 遗传进化 故障字典 诊断系统模型 

分 类 号：TN79[电子电信—电路与系统] TN707