Chaotic Neural Network-based Analog Circuit Fault Diagnosis

The analog fault diagnosis research already formed a series of diagnoses theories and methods since 20th century 60's. However, complicated diagnosis question makes these methods's validity poor. A new effective approach is provided based on AI(artificial intelligence) like neural network. In this paper, a method of analog fault diagnosis is researched based on chaotic neural network (CNN).One important aspect of analog fault diagnosis is ascertaining and testing to analog fault, which is discussed in this paper based on chaotic neural network, the method of establishing fault dictionary being the same with CNN is also researched. Arithmetic of variance data inosculation is discussed to process original data, compressing and processing to the original data is realized by using this arithmetic.As a new dynamics theory, CNN has been extensively applied to model forecasting, pattern identification, fault diagnosis field et al, and is gradually regarded by many scholars in several fields. A method of fault diagnosis to liner analog is researched based on improved Aihara CNN in this paper, compared with traditional Back Propagation (BP) network and momentum BP network is also discussed. A method of fault diagnosis to nonlinear filter analog is researched based on improved NCNN (Noise Chaotic Neural Network), the problems are discussed happened during data processing and network testing, and the effect of diagnosis is validated.After doing a mass of emulation experiment, the analog fault diagnosis based on CNN can speedily confirm fault model and exactly realize fault forecast. Compared with traditional network, the method has strongpoint of high diagnosis ratio, short diagnosis time and strong illegibility ability. In conclusion, the algorithm has improved the robustness and practicality of analog fault diagnosis by AI.