摘要
文章提出了一种基于小波神经网络的模拟电路故障诊断方法:通过分析被测电路的冲激响应来识别电路中的故障元件,利用小波理论中的多分辨率分析的方法提取出相应信号中的故障特征,组成特征向量后输入神经网络进行训练,实现故障诊断;该方法减少了神经网络的输入、简化了其结构、并缩短了训练和处理时间,文中分别用小波神经网络和传统的BP神经网络对实例电路进行故障诊断,仿真结果发现:小波神经网络相比BP网络方法收敛速度更快,诊断率更高。
A method of analog circuit fault diagnosis based on wavelet neural network is presented in this paper.Faulty components were identified by analyzing impulse response of circuit under Test.According to the multi-resolution analysis method in wavelet theory,the feature information is extracted from the impulse response.Then put the feature vector to network for training,and the diagnosis is realized.This method reduces the number of inputs to neural network,simplifies its architecture and minimizes training and processing time.The wavelet neutral network and traditional BP network are used respectively in the example circuit for diagnosis.Simulation result indicate that the diagnosis rate of wavelet neural network was higher than BP network,and the convergence speed is fast than BP network.
出处
《计算机测量与控制》
北大核心
2014年第11期3521-3524,共4页
Computer Measurement &Control
关键词
模拟电路
故障诊断
小波分析
神经网络
Analog Circuit
Fault Diagnosis
Wavelet analysis
Neural Network