摘要
为了解决容差模拟电路故障诊断,采用了一种集成T S模糊神经网络的故障诊断方法,首先应用PSpice软件仿真得到故障数据,其次对故障数据进行小波分解并归一化处理,获得神经网络训练样本,然后将样本分配到每个T S模糊神经网络中进行训练和测试。采用学习速率可变的附加动量BP算法训练网络权值,使其稳定性与收敛速度达到最佳。仿真结果证明:该方法收敛速度快、正确诊断率高,能有效地实现对模拟电路故障的诊断。
In order to solve the problem of tolerance analog circuit fault diagnosis, a fault diagnosis method based on an in- tegrated T-S fuzzy neural network was used. PSPICE software simulations is applied to get fault data, and then the wavelet de- composition and normalization of fault data are executed to obtain the neural network training samples. Finally, the samples are assigned to each T-S fuzzy neural network for training and testing. In the training process, the additional momentum BP algo- rithm, whose learning rate is variable, is used to train the network weight for making its stability and convergence speed best. The simulation results show that the approach has fast convergence and high accuracy. It can effectively realize the analog cir- cuits fault diagnosis.
出处
《现代电子技术》
2013年第4期133-135,140,共4页
Modern Electronics Technique
基金
海南省自然科学基金资助项目(610231)
关键词
模糊神经网络
模拟电路故障诊断
集成神经网络
学习速率可变
fuzzy neural network
analog circuit fault diagnosis
integrated neural network
learning rate variable