Abstract: The research on theory and method of analog circuit fault diagnosis is a hot research topic at present. Traditional neural network learning algorithm has limitations, such as it is difficult to determine its structure, to comprehend and to achieve with the hardware. To solve these problems of traditional neural network in fault diagnosis of analog circuits, we used the neighborhood covering algorithm (NCA) to determine the structure of the neural network. In order to reduce number of neurons and determine a good initial point of NCA, this paper studied an improved method (NSCA). In the end, the fault diagnosis of a certain band-pass filter circuit, which reduces the number of neurons in the neural network, improves the generalization ability of the neural network, and improves the accuracy of diagnosis about 9 percentage point. Simulation results show that the method is more effective.

Key words: analog circuit, fault diagnosis, neighborhood covering algorithm, neighborhood search covering algorithm