Abstract: In order to solve the problems caused by large dataset，such as the network scale and the training time，distributed diagnosis algorithm for analog circuit fault is presented in this paper.This algorithm adopts supervised Hebb learning rules.In addition，because the class flag is added to the algorithm，the learning algorithm doesn’t discard the partial information due to data partitioning.This distributed algorithm and traditional BP algorithm are used respectively in the fault diagnosis of and example circuit.Simulation results indicate that the distributed algorithm in this paper not only has a equivalent accuracy with BP algorithm，but also speed up the learning rate effectively.

Key words: distributed neural network, analog circuit, fault diagnosis, supervised learning

摘要： 提出一种模拟电路故障的分布式诊断算法，用以解决大数据量故障样本集所带来的网络规模过大，训练时间过长等问题。该算法采用有监督Hebb学习规则，在训练学习过程中添加类别标识，避免了因数据分割而产生的部分知识的丢失。分别用提出的分布式算法和传统的BP算法对实例电路进行故障诊断，实验结果表明，提出的分布式算法不仅和BP算法的诊断正确率相当，而且有效地提高了训练学习的速度。

关键词: 分布式神经网络, 模拟电路, 故障诊断, 有监督学习