摘要
针对目前神经网络模拟电路故障诊断中存在的难点,提出了基于信息融合思想的多神经网络故障诊断方法;该方法测试电路中节点电压信号、供电电流信号,利用小波变换对检测信号进行预处理,基于主成分分析对特征矢量进行降维,根据模拟电路的不同故障模式分别建立诊断神经网络,用概率统计数据融合方法从多个神经网络中选出最优网络用于诊断故障;通过电路实例验证了新故障诊断方法的有效性,实验结果表明新方法可有效提高故障诊断性能。
A novel method of analog circuit fault diagnostic based on neural network with wavelet transform and information fusion＇are presented, The node voltage and supply current are tested in the method. Wavelet decomposition is used as preprocessor for fault information. Using principle composition analysis reduces the dimensionality of fault features. Networks with the highest prior probability are selected for future fault diagnosis. It＇s shown by an example circuit that the novel method can effectively promote the performance of fault diagnosis.
出处
《计算机测量与控制》
CSCD
2006年第2期146-148,共3页
Computer Measurement &Control
基金
航空科学基金资助项目(04I52068)
关键词
模拟电路
故障诊断
小波变换
神经网络
信息融合
analog circuits
fault diagnosis
neural network
wavelet transform
information fusion
