Abstract: This paper researches the method based on the wavelet multi-resolution feature extraction for analog circuit fault diagnosis. The features of sampling signal are extracted from wavelet decomposition as a vector to BP neural network for training. According to bandpass filter circuit of diagnosis example, this paper discusses the specific implementation of this method. It can effectively simplify neural network structure and reduce its training time, quickly and efficiently carry out analog circuit fault diagnosis and positioning.


Key words: wavelet analysis, neural network, analogue circuits, fault diagnosis