基于空间特征的非线性电路神经网络诊断  

Neural Network Fault Diagnosis in Nonlinear Circuits Based on Space Characteristic

作　　者：龙春阳[1] 谭阳红[1] 

机构地区：[1]湖南大学电气与信息工程学院

出　　处：《电气电子教学学报》2008年第2期63-66,共4页Journal of Electrical and Electronic Education

摘　　要：本文提出线性模拟电路的单、双、三故障空间特征,采用分段线性模型(PWL)将非线性电路线性化,通过遗传算法求电路的容差范围,用神经网络对非线性网络进行诊断。本文的方法大大减少了模拟计算量,同时,使神经网络的训练过程加快,训练误差减少,并大大提高了诊断的正确率。This paper presents the single , double and three fault space characteristics of the line - analog circuits. Nonlinear circuit can be transformed line circuit by adopting a cent segment line model （ PWL） . Genetic algorithm to calculate tolerance scope of circuit is given, and using a neural network to diagnose nonlinear network using a neural network. The method consumedly reduces simulated calculation quantity, makes the training process of neural network speed in the meantime, trains error margin decrease, and consumedly raises a diagnostic correctness.

关 键 词：空间特征 分段线性模型(PWL) 非线性电路 神经网络 

分 类 号：TM132[电气工程—电工理论与新技术]