The model of the state diagnosis for complex system based on the improved fuzzy neural network

Aiming to deal with the issue that troubles the knowledge accumulation in a lot of complicated systems in the form of expertise that is showed as the fuzzy language, and fuzzy neural network (FNN) based on the traditional fuzzy algorithm is difficult to utilize expertise to get abundant training samples directly, this thesis proposes the FNN model based on the improved fuzzy algorithm. In this model, the improved fuzzy algorithm is not only used to deal with the input amount, but also to directly change the expertise into training sample that is necessary in neural network training. Then, a FNN diagnosis model based on the improved fuzzy algorithm is put forward. Compared it with the model based on the traditional fuzzy algorithm, the model designed in this thesis proves to be more effective, more clearly reasoned and more quickly in analysis. Moreover, it can produce sufficient training samples. All of these advantages will be useful to the accumulation of experience, online training and the improvement of FNN diagnosis precision.