摘要
本文提出了利用故障信号各频带的系数序列的绝对值之和,和能量特征作为特征信号经神经网络信息融合的模拟电路故障诊断新方法。该方法先对采样后的故障信号进行小波分解,提取故障特征信号经归一化和向量关联后作为特征向量输入BP神经网络进行训练诊断。通过电路诊断实例,阐述了该方法的具体实现,验证了所提方法的有效性。
A new method for fault diagnosis of analogue circuits based on neural network information fusion technique about the sum of absolute values of coefficients sequence and energy features in each frequency band of fault signals was put forward. The method was firstly to make wavelet decomposition for the fault signals after sampling, then to extract fault characteristic signals to the normalization and take the relevance vectors as the feature vectors to input BP neural network so as to make a training diagnosis. Through the example of circuit diagnostics, the paper expounded the concrete realization of the method and demonstrated the effectiveness of the proposed method.
出处
《科教导刊》
2013年第13期190-192,共3页
The Guide Of Science & Education
关键词
小波分析
神经网络
模拟电路
故障诊断
wavelet analysis
neural network
analogue circuits
fault diagnosis
