Abstract: In this paper,a fault diagnosis method based on Haar wavelet and BP neural network for nonlinear analog circuit has been presented.The proposed method by using wavelet decomposition as a preprocessor of nonlinear analog circuit fault signal can drastically reduces the number of inputs and training and processing time of the neural network.After briefly introducing an improved BPNN which adpots momentum factor for preventing local convergency used for nonlinear analog-circuit fault diagnosis,the basic principle based on Haar wavelet decomposition that can extracts the fault feature from fault signal is discussed.

Key words: HAAR wavelet transform, neural network, nonlinear analog-circuit, fault diagnosis, preprocessor