Research On Fault Diagnosis Technology Of Wireless Sensor Network

In recent years,network technology has been booming,and Wireless Sensor Networks(WSN)has attracted attention as one of the representatives of the Internet of Things.Since the deployment of WSN,each node has been almost unattended and once it fails,it may cause huge economic losses to users.Therefore,timely diagnosis and elimination of WSN failures and improving the reliability and life cycle of the WSN network operation are the preconditions for ensuring that the WSN monitoring system completes the specified tasks.According to the structural characteristics and functional characteristics of WSN,this thesis presents a fault eigenvalue extraction method and sample data partitioning method for the possible faults in WSN.A PNN-based WSN fault diagnosis method and a SVM-based WSN fault diagnosis method are proposed.And compared with the traditional BP algorithm simulation analysis,the results show that they have a higher reliability.In addition,this thesis also designs and develops the WSN fault diagnosis system.The specific content includes the following aspects:(1)The common failures of WSN are studied,including the types of failures and the hazards caused by the failures.The WSN failures are analyzed to determine the types of WSN failures to be studied in this thesis and the characteristics of failures at the time of WSN failures,and between the failure characteristics and the failure types.The correspondence.(2)Considering that different fault features have different effects on the fault diagnosis results,the MIV algorithm is used to screen and reduce the fault features before the simulation,and the fault features with large influence factors on the simulation results are extracted.Considering that the spatial distribution of sample data may be irrational,the SPXY algorithm is used to divide the sample data so that the final training set data and test set data can reflect the characteristics of the sample database data.(3)Several common traditional algorithms in fault diagnosis of WSN are studied,including BP neural network,RBF neural network and Bayesian network.It is found through analysis that these traditional algorithms still have low diagnostic accuracy and unstable network in WSN fault diagnosis.Such disadvantages,based on this thesis proposed based on PNN-based WSN fault diagnosis model and SVMbased WSN fault diagnosis model.Compared with the BP network simulation results,the results show that the proposed two algorithms have higher reliability in WSN fault diagnosis.(4)Using Web development technology,a WSN fault detection system was designed,including data visualization of WSN fault feature values,WSN system communication status assessment,and WSN system fault analysis and reporting.