Research And Implementation Of WSN Node Fault Intelligent Detection Method Based On Sensing Data

With the rapid development of embedded applications and microelectronics technology,Wireless Sensor Network(WSN)technology is based on people's needs are applied in a variety of occasions.Because WSN which would be generally deployed in harsh and unattended areas of the environment,nodes maybe perceived failure or damage,which the failure of sensor module is particularly common.There are some problems which are solved in the traditional algorithm,such as,fault detection accuracy is low,fault detection alarm rate is high,and excessive reliance on the experience.This paper studied the above and the experimental verification is carried out in the practical application,the research as the followings:Firstly,research on node fault detection method which based on perceived data in WSN.For the traditional algorithm,there is not enough ability to characterize the perceived data and consider the spatial and temporal correlation,fault detection accuracy is low.This paper presents a node fault detection method which based on perceptual data.The innovation is to establish the data model of multidimensional perceptual data using the spatiotemporal correlation of perceptual data.The model includes two parts,Firstly,it has established the multidimensional space vector for the perceptual data acquired by each node,then the cross window is constructed,and the historical data is introduced from the time-dependent perspective as the horizontal window,the sensing data from neighbor nodes is introduced from the spatial correlation angle as the vertical window.Then,the node detection mechanism is used to determine the fault node.The Experiments show that the accuracy of fault detection is improved by 15.5% and the false alarm rate is reduced by 4.89%.Secondly,research on sensor fault detection which is intelligent based on Artificial Neural Network.In the view of the traditional algorithm,non-intelligent,inaccurate to determine the status of the sensor module and excessive reliance on the experience.This paper has proposed a fault detection method of sensor module based on Artificial Neural Network which based on node fault detection method that based on perceptual data.In order to select the most suitable artificial neural network,this paper selects the BP,RBF and Hopfield neural networks to compare the experiment.Experimental results show that BP neural network has better recognition effect in sensor fault detection than other artificial neural networks.The fault detection module based on BP neural network not only improves the accuracy of fault detection module detection,but also solves the problem that the threshold of fault detection module is too dependent on expert experience.Lastly,design and implementation of WSN Intelligent Fault Detection Prototype System(IFDPS).It realizes the node fault detection method based on the spatiotemporal correlation and the intelligent fault detection method of the node sensor module based on Artificial Neural Network,including the management of data and nodes,fault detection module and other modules,and proves the validity of the theory in the practical application.