With the rapid development of the Electronic Industry, the importance of the analog circuit fault diagnosis is more and more obvious, it has important signification for working orderly and dependability design of electronic equipment or system. The traditional methods of fault diagnosis are performed only if the faults of the circuits are those hard faults, such as open-circuit, short-circuit, etc. Those soft faults aroused by the tolerance of circuit components cannot be easily discovered. However, a method based on neural network can easily solve these problems.For the neural network has the function of processing complex multi-patterns and carrying on the association, the extrapolation and the memory, it is particularly suitable for fault diagnosis system. Fault diagnosis by applying the neural network in analog circuit has become the most recent development tendencyćIn this thesis, development, actuality and potential problems of the analog circuit fault diagnosis are discussed firstly, introduced BP neural network model architecture and studying mechanism. Several fast learning algorithms for the BP networks were analyzed. To further improve the speed of convergence and avoid the shortcomings of sticking to a local minimum, a new algorithm based on Lyapunov stability was presented here. It has been proved to solve this issue effectively.In the thesis, some fault feature extraction methods are studied, including effective sampling point extraction and wavelet analysis. A typical filter circuit is selected. Using the BP network based on improved heuristic algorithm and Lyapunov theory respectively, the fault of the chosen typical filter circuit is diagnosed. The simulation results of the example show that the fault diagnosis methods based on the neural network have good diagnosis effect and feasibility in tolerant circuit.In this thesis, some research results are obtained, but those are not easily applied to the practical application due to the difference from theory to the practice, and it still should be studied in the future. |