Fault Diagnosis Model of WSN Based on Rough Set and Neural Network Ensemble

An intelligent fault diagnosis model of wireless sensor networks (WSN) using rough set and artificial neural network ensemble (RS-ANNE) is developed to solve the fault diagnosis problems of WSN such as limited energy and substantive information redundancy, thus prolonging service life of the whole WSN effectively. The attribute reduction for decision of fault diagnosis is utilized based on the discriminate matrix in rough set theory. The minimum fault diagnostic characteristics subset with the greatest contributions is selected so that preliminary topological structure of the neural network is determined. The network is trained to reflect mapping relationship between inputs and outputs, and network ensemble is used to realize the fault diagnosis. Simulation results show that diagnostic accuracy of the proposed method is 95.67%. Computation amount of RS-ANNE is decreased by 13.88% and diagnosis accuracy is increased by 22.98%, compared with those of ANNE.