Improved Fault Diagnosis Method Based on Probabilistic Neural Network

Abstract

In order to enhancing the accuracy of fault diagnosis system, an improved method based on the probabilistic neural network (PNN) is proposed, in which the synthetic attribute weights of faults are introduced that are obtained by integrating algebra view and information theory view of rough set. The synthetic attribute weights are utilized to training the classical PNN and dealing with the classification of faults so as to improving the PNN model. The new model is more accurate and can represent expertise. This novel approach is applied in digital data network to diagnose failures, and the results of the experiment verify that the method is practical and effective in raising accuracy of diagnosis as well as avoiding misdirection in fault remedy.