Granular Computing and Neural Network Integrated Algorithm Applied in Fault Diagnosis

Abstract

A granular computing and neural network integrated algorithm is applied in fault diagnosis, taking advantage of the knowledge reduction ability of granular computing and good classified diagnosis ability of neural network. After data acquisition and pretreatment, the fault samples are discreted to form a decision table. The attributes reduction based on binary granular matrix can find minimum attribute set under the same classification ability. And then the reduced system is utilized to the neural fault classifier, where granular-computing-based-reduction reduces the dimension of input to neural network and improves the efficiency of training. A fault diagnosis example of the hydrogenerator unit shows the effectiveness of the proposed method in the paper.