Design Of Fault Diagnosis Algorithm In Wireless Sensor Networks

Due to the unpredictably complex working environment of wireless sensornetworks, it is easy to cause the sensor nodes to be faulty. Fault diagnosis is to detectabnormal behaviors of nodes using a certain mechanism, and make a diagnosis result.Therefore, fault diagnosis has very important significance to manage the reliability ofwireless sensor networks. Fault diagnosis methods in wireless sensor network usingclustering and neural network are respectively studied in this paper as follows:Firstly, this paper proposes a distributed clustering fault diagnosis method todiagnose the intermittent faulty sensors of sensor nodes. The nodes subjected tointermittent faults sometimes behave as fault-free. First, the network is divided intoseveral clusters by diagnosing with the cluster heads. The selected cluster heads shouldnot only meet the conditions of the cluster heads, but also be diagnosed to be good.Then, the cluster members are diagnosed by their cluster head in every cluster. In thispaper, the strategy of improving the diagnostic accuracy of intermittent faults is sensingand analyzing data in several times. The strategy is explained to be diagnosable, and theresults of the theoretical analysis of fault diagnosis accuracy are given. The simulationresults show that the proposed algorithm has high fault detection accuracy.Secondly, this paper proposes a diagnosis method of nodes in wireless sensornetwork using RBF neural network. First, make reduction of fault diagnosis informationof nodes in wireless sensor network with rough set. Then use a set of training samples totrain the RBF network. Finally, diagnose the testing samples using the trained RBFnetwork. In the simulation, RBF network is compared with BP network which has beenused in most diagnosis methods, and the results show that the training speed of RBFnetwork is much higher than that of BP network, and without considering faultinformation accuracy, data transmission errors and other factors, the fault diagnosismethods using these two networks both have high fault diagnosis accuracy.