Application And Research On Information Fusion Technology In The Field Of System Fault Diagnosis

Information fusion is a comprehensive intelligent technology. Fault diagnosis has the same aim and demand with it in principle. Moreover, multi-sensor information fusion could provide more information to fault diagnosis which is used to analyze and composite fault information to conduct the diagnosis results that are more legitimated, comprehensive and accurate. Therefore, combining information fusion with system fault diagnosis has important meaning.Firstly, this thesis researched on information fusion and analyzed principle and procedure of fault diagnosis. Then, fuzzy neural network and D-S evidence theory and its applications in both fields were explored. Finally, considering the problems such as the low recognition rate and diagnosis precision, this thesis presented and constructed a new model of information fusion and fault diagnosis according to fault features of current systems. This model was constructed by feature and decision levels of fusion diagnosis. The concept, fusion diagnosis center based on fuzzy neural net theory, was presented in feature level fusion diagnosis. And then some methods were found to improve the deficiencies of hybrid learning algorithm like the low convergence speed and the concussion. The effect of convergence was better than original algorithm. At the same time, the efficiency of the whole fusion diagnosis center was increased. The model was designed to adapt to diverse fusion diagnosis centers according to different fault features of systems, and we used D-S evidence theory to process the global fusion diagnosis of decision level. In the end, we did some experiments via MATLAB and proved the feasibilities and effectiveness of the model. The recognition rate and the diagnosis precision were improved, which were found from experiments.