无线传感器网络故障检测机制的研究

With the wide application of Wireless Sensor Networks (WSNs) in production and living, fault detection in WSNs has become a research focus due to ensuring healthy operations of nodes and network. Based on a large number of surveys on fault detection and location of WSNs, different fault detection models and mechanisms are built in the project through the classification of fault types, node types and network types. For data drift caused by sensing fault, the uncertainty-based distributed fault detection mechanism and multi-variable compression-based fault detection mechanism of WSNs are proposed and improve the accuracy and efficiency of fault detection in homogeneous and heterogeneous networks. For communication fault of links and nodes, the project firstly researches on the transmission strategy of active probes based on maximum prior probability. And then a static Bayesian network algorithm with low complexity is built by fault region division to reason the maximum probable set of fault links. Meanwhile, the dynamic Bayesian network mechanism is proposed to improve the accuracy of detection on node communication fault in dynamic topology networks. The mechanisms of fault detection proposed in this project, which are well targeted, real-time and archive high accuracy and efficiency for detection, are of great theoretical significance and realistic demand for the researches on WSNs..

随着无线传感网络在生产生活中的广泛应用，对其进行故障检测以实现节点和网络的正常运行成为目前的研究热点。本项目在对无线传感网络故障检测和定位机制进行大量调研的基础上，根据不同的故障类型、节点类型和网络类型来构建相应的故障检测模型和机制。针对传感器的数据感知故障所带来的数据漂移问题，提出基于不确定性的分布式协同故障检测和基于多变量压缩的异构数据故障检测，提高在同构及异构网络环境中故障检测的精度和效率；针对链路通信故障和节点通信故障，研究基于最大先验概率的主动探测发送策略，通过故障区域划分构建低复杂度的静态贝叶斯模型实现对链路故障集合的推理，并提出动态贝叶斯网络模型提高在网络动态拓扑环境下对节点通信故障的检测效果。本项目所研究的故障检测机制针对性强、检测精度高、检测效率高、实时性强，对于无线传感器网络的研究有着重要的理论意义和现实需求。

随着无线传感器技术在我国各个城市的不断推广，因无线传感器故障所引起的各种负面效应也逐步凸显。例如，传感器感知模块发生故障时将引起数据漂移和恒增益现象，从而影响数据的精确融合，干扰用户的正确判断，在智能医疗领域甚至还会酿成重大的医疗事故。传感器故障检测机制的研究对无线传感器网络的正常运行有着重要的理论意义和现实需求。.首先，本项目针对传感器的数据感知故障所带来的数据漂移问题，研究基于数据聚类及异常值检测的数据故障诊断机制；研究基于时空关联性的通信故障检测机制和基于贝叶斯网络的链路故障推理机制，提高在无线传感器网络动态拓扑环境下对节点通信故障的检测效果。在研究期内，上述研究内容满足任务书要求。.其次，为了对故障节点进行高效识别和准确检测，项目还扩展研究了基于可靠性的传感器数据传输和高效存取技术，包括基于备份路径的数据传输协议、基于预测的传感器网络中继选择方法、基于可靠性的传感器数据文件存放策略等。.再次，对已经形成网络故障后的无线传感器网络连通性恢复方法进行了探索性研究，提出了考虑障碍的网络分区连通性恢复算法，该算法研究移动中继节点混合放置策略，从而恢复网络连通性；提出了异构无线传感器网络容错拓扑控制方法，构建普通节点通向超级节点的多路由树，确保异构无线传感器网络的故障容忍。.最后，基于本项目的理论研究成果，项目组还研发了应用于集装箱堆场环境下的无线传感器网络监控和故障诊断平台。该平台可以自动采集数据、拓扑展示、历史数据查询、识别故障并定位。通过在实际集装箱堆场环境中进行部署，可对集装箱货物尤其是危险品运输提供技术支撑，避免因天气变化造成危险货物运输和保管过程不当造成的严重危害。.本项目围绕故障的感知、检测以及基于故障的网络连通性恢复等一系列流程进行研究，圆满并超额完成了项目任务。