Research On Fault Diagnosis Methods Of Wireless Sensor Networks

As the wireless sensor networks(WSN) are more and more widely used in all kinds of monitoring system, the research of it is increasingly more important. After the deployment of the wireless sensor network nodes, the nodes are in the condition of unmanned monitoring and inspection. We don’t know the running state of the nodes and it’s impossible to monitor them in real-time or often check them. If the sensor networks’ faults happen, the faults may influence the motoring tasks of the networks. Therefore, diagnosing the faults of wireless sensor network nodes accurately and timely can improve the reliability of the wireless sensor networks’ operation and ensure the monitoring systems which used WSN to accomplish the monitoring tasks. The paper makes a further study in the fault diagnosis methods of wiressless sensor networks. The main research contents are as follows:(1)Studied the rough set theory and maked a decision table through the WSN node fault types and the corrspoding fault feature attributes. Then used the rouset theory to reduce the attributes in the WSN fault decision table and did a simulation experiment on the WSN fault diagonosis method which based on roughest theory. The results showed the superiority and deficiency of the method.(2)Studied the wavelet neural network which based on BP algorithm. As this WNN usually has low convergence rate and easily falls into local minimum which is due to the using of gradient algorithm, an improved wavelet neural networks is proposed. In the improved wavelet neural networks, the momentum coefficient and alter-learning coeffient are employed to resolve the problems above. The reliability of this improvement measures was proved through the training experiment. At last, the improved wavelet neural network algorithm was used in WSN fault diagnosis exeperiment. The exeperiment not only proved the algorithm’s feasibility in WSN fault diagnosis, but also showed the algorithm’s good fault tolerant performance.(3)As the WSN fault diagnosis method based on rough set is lack of fault tolerance fault and the wavelet neural network can not recognize the extra data, we proposed a fault diagnosis algorithm which intergrates the rough set theory and the improved improved wavelet neural network to solve the problem. Then we did a WSN node fault diagnosis experiment with the RS-IWNN fault diagnosis algorithm. Compared with the experiment results of WSN fault diagnosi method based on rough set theory, the superiority of RS-IWNN fault diagnosis algorithm is proved.(4)As the DFD algorithm has the problems of high energy consumption and diagnosing the node as normal in a harsh condition, we improved the DFD fault diagnosis algorithm when the WSN runs the CTP agreement. With the improved DFD algorithm, we can reduce the energy comsumption of the WSN fault diagnosis and improve the diagnosis accuracy. The experiment showed the good results the algorithm achieved.(5)For the current study mostly focuses on the WSN fault diagnosis algorithm and ignores the design of the diagnosis system, we proposed a design of WSN fault diagnosi system and introduced the design elements.