摘要
以近年来国内外有关的文献报道为依据,对目前已经提出的各种基于神经网络的模拟电路故障诊断方法进行系统的归纳和分类,重点讨论了神经网络故障字典法和神经网络优化诊断法;指出模拟电路故障诊断的神经网络诊断法不能完全取代传统的诊断方法,并预测这类方法的发展趋势是应用小波变换、模糊控制和遗传算法等技术,克服神经网络本身的局限性,并解决神经网络结构的确定、数据预处理和训练样本集的优选等问题。
A summary of the neural network approach to analog circuit fault diagnosis is given by analyzing the concerned literatures published in recent years. Every neural network approach is summarized and classified by its theory. The neural network fault dictionary approach and neural network optimization diagnosis approach are discussed as emphases. It is pointed out that the neural network approach could not substitute for classical approach to analog circuit fault diagnosis absolutely. The research trend of this area is the application of technologies such as wavelet transform, fuzzy logic and genetic algorithms to overcome the disadvantage of neural network itself, fix on the structure of neural network, preprocess data and select the training sample sets.
出处
《计算机测量与控制》
CSCD
2006年第5期564-566,共3页
Computer Measurement &Control
