Research of condenser fault diagnosis method based on neural network and information fusion
Xia Fei, Hao Zhang, Zhang Kai, Peng Daogang · 2010
According to the insufficiencies in condenser fault diagnosis based on single neural network, a new method of condenser fault diagnosis based on neural network and information fusion has been proposed this paper. By means of grouping fault symptoms, the different neural networks have been adopted to diagnose faults. And the results are composed of the preliminary fault diagnosis synthetic matrix. In order to combine the diagnosis results of each neural network, this paper has focused on discussing the way to confirm confidence degree of each neural network. The weight matrix during the process of information fusion has been made up of these confidence degrees, which is calculated with the preliminary fault diagnosis synthetic matrix to finish the fusion of several diagnosis networks. During the stimulation test of the condenser faults, the method presented in this paper has a higher accuracy than that of traditional neural network method. Especially the probability of unrecognized fault type has been reduced in condenser fault diagnosis.