Abstract:

In the use of BP neural network to diagnose fault in analog circuits,the network input that represents fault signature is very important.The common characteristics of information structure and fault diagnosis method are introduced,and a new method based on multi-test point multi-feature information of the original sample set is proposed.The original fault signature set is constructed as the input of BP neural network to train the network.Simulation results show that the network trained with sample set by this method offers better accuracy than those by traditional methods in fault diagnosis of analog circuits.This novel method for fault diagnosis of analog circuits proves feasible.

Key words: BP neural network;analog circuits;fault diagnosis;fault feature