摘要
提出了一种基于模拟电路故障诊断的神经网络方法。这种方法利用小波分解、数据标准化、主成分分析对输入数据进行预处理,采用k个神经元输出的前馈神经网络结构进行有效训练。该方法检测和识别故障准确率高,系统的鲁棒性和稳定性强。
An analog circuit fault diagnosis method based on neural-network is presented. The method uses wavelet decomposition and principle component analysis to preprecess input data. The fault diagnosis system is based on 1-k neural network architecture using bavk-propagation algorithm for training. The system has capability to accurately detect and identify fault components in an experiment circuit. And it provides a more robust and stable fault diagnosis.
出处
《电讯技术》
北大核心
2004年第3期43-46,共4页
Telecommunication Engineering
基金
吉林省自然科学基金项目"集成电路及器件故障诊断方法及应用研究"(20010582)
关键词
模拟电路
故障诊断
神经网络结构
主成分分析
Analog circuit
Fault diagnosis
Neural-network architectures
Principle component analysis
