Node Fault Diagnosis Algorithm On Wireless Sensor Networks

Wireless sensor network is a wireless network composed of a large number of static or mobile wireless sensor nodes through multiple hops in a self-organizing manner,which is mainly used for data sensing,acquisition,processing and transmission.Wireless sensor nodes are widely deployed to harsh environment or inaccessible areas for information monitoring because of their small size and self-organization to form a network,so sensor nodes are prone to a variety of failures so that the perceptual data is not sent to the base station for analysis or perceptual data error,resulting in errors in the subsequent analysis,Which leads to inestimable consequences.Therefore,it is of great importance to diagnose the fault of sensor nodes.In this thesis,different fault diagnosis algorithms are proposed for different scenarios and optimization requirements in the network.(1)To reduce the low fault diagnosis accuracy rate of traditional distributed fault diagnosis algorithm with the high node failure rate in wireless sensor networks,a fault diagnosis algorithm for wireless sensor networks based on neighbor node pre-state is proposed.The algorithm integrates the sensing data of neighbor node and sensor node to diagnose the sensor nodes' state,and transmits it to the base station by the mobile charging sensor node,which improves the fault diagnosis accuracy without increasing the energy consumption.This algorithm has high fault diagnosis accuracy rate in a network with high node failure rate,and has a good application prospect.The accuracy of fault diagnosis is improved,and the algorithm can still achieve high accuracy of fault diagnosis in networks with high failure rate.(2)To reduce the energy consumption in node fault diagnosis with high node failure rate in WSNs,a fault diagnosis algorithm based on the mobile charging sensor is proposed.The proposed algorithm clusters the sensor nodes according to the minimum uniform clustering algorithm,firstly.Then,the cluster heads diagnose the sensor nodes according to the similarity of the neighbor nodes.Finally,the cluster heads are charged by the mobile charging sensor which collect the state of sensor nodes.The proposed algorithm can greatly reduce the energy consumption in the network,ensure the accuracy rate of fault diagnosis,and prolong the service life of WSNs.(3)To reduce the high energy consumption and the accuracy of fault diagnosis decreases with the increase of node failure rate of traditional fault diagnosis algorithm with the high node failure rate in the sparse network,a convolutional neural network based on mobile charging sensor is proposed.The proposed algorithm introduces a mobile charging sensor to transmit sensing data to the base station and charge the sensor nodes that need to be charged,and then the base station trains the convolutional neural network to diagnose sensor nodes.The fault diagnosis accuracy of the algorithm does not change with the failure rate,and it can guarantee the lifetime of the wireless sensor network and can classify fault node,meanwhile it can prolong the service life of the network.