The Research On Data Fault Tolerance In Wireless Sensor Networks

Wireless Sensor Networks (WSN) is a data-centered distributed sensor network, which is widely used in many fields, including military, environmental monitoring, health care and so on. It’s an important research direction at home and abroad. However, in the practical application of WSN, the selected data may be inaccurate or even false and the service quality cannot be guaranteed due to many factors. For example, nodes with limited energy are usually deployed in harsh environments, which may cause high failure rate, and wireless communications are easily jammed or affected by other unfavorable factors. Therefore, WSN system should have the corresponding data fault tolerance to ensure the accuracy and validity of the data obtained.The main research objects of WSN data fault tolerance is the false data caused by adverse factors of the system. We can reduce the impact of false data by fault detect and data aggregation and then to obtain the accurate values. Based on the previous fault-tolerant algorithm, this thesis made a further investigation of node failure detection algorithms, data aggregation algorithm and data fault-tolerance mechanism, and proposed a node failure detection algorithm based on event-driven, a data aggregation fault-tolerance mechanism based on event clusters and a K-Means clustering center selection algorithm based on domain overlap ratio.The algorithms and mechanisms proposed in this thesis made several innovations:Firstly, the specificity of quasi event boundaries and the node with insufficient spatial redundancy information is taken into account in the process of node failure detection; the correlation detection algorithm based on time is adopted to improve the detection accuracy. Secondly, this thesis proposed K-Means algorithm based on domain overlap ratio:using the rate factor of overlapping data object fields, we can obtain more accurate data clustering center value through the algorithms in the operation of low-dimensional data aggregation. Thirdly, a data aggregation fault-tolerance mechanism is constructed based on multiple event clusters. The node reliability feedback mechanism is designed to determine and isolation the false node; and then using K-Means algorithm based on domain overlap ratio to realize data aggregation for each cluster. The result of simulation experiment shows that the related algorithm and mechanism proposed in this thesis improves the accuracy of node failure detection; reduce the amount of data transmission; improve data aggregation effect; and reduce the relative error of convergence results.