基于局部决策融合的无线传感器网络诊断方法研究

Fault diagnosis is an important factor to improve the robustness of wireless sensor networks. In large scale sensor networks,the sink-based diagnosis methods cause large overhead and always have difficulty in getting the network status from the problematic area.To overcome these disadvantages,a local evidence fusion diagnosis scheme for wireless sensor network is proposed in this project.The diagnosis scheme tries to study the relationship between the network status and the fault causes.Then,it focuses on an evidence fusion model which is used to detect the network faults and to infer the root-cause of faults.The research include three parts. In the first part, a data feature extraction method based on statistic correlation sequence is first to be studied, then the correlaton models between the network status and the faluts are to be proposed. In the second part, to achieve consistent diagnosis results, an improved D-S theory is to be used to conduct evidence fusion.Finally, in the third part, the fault dection scheme for the local evidence fusion diagnosis is to be studied in detail. In order to improve the diagnoses efficiency, appropriate fusion algorithms based on fusion-tree should also be proposed. Compared with sink-based diagnoses,the local evidence fusion diagnosis needs only information from local nodes.It can get consistent and correct diagnosis results while reducing the diagnosis overhead.

开展无线传感器网络故障诊断技术研究对提高传感器网络的可用性具有重要的意义。在大规模的无线传感器网络中，传统的集中式故障诊断方法存在诊断开销大、局部故障区域状态获取不准确等不足。本课题致力于研究一种基于局部决策融合机制的传感器网络诊断方法，课题拟通过研究局部节点状态信息与网络故障之间的关联性，通过采用高效的决策融合模型来检测和推理网络中出现的异常和导致故障的原因。研究围绕以下几方面展开：1）研究基于统计关联排序机制的节点状态特征提取方法，并建立状态信息与网络故障的关联模型；2）研究基于改进D-S理论的节点诊断证据融合模型，实现一致性的局部故障诊断；3）研究局部诊断的故障发现机制，通过设计基于融合树的决策融合算法，提高故障诊断的效率。与传统的传感器网络诊断技术相比，基于局部决策融合的故障诊断方法仅需要局部节点的状态信息，不仅可以有效降低故障诊断的开销，还能够在局部得到一致而又准确的诊断结果。

在大规模的无线传感器网络中，传统的集中式故障诊断方法存在诊断开销大、局部故障区域状态获取不准确等不足。本课题研究了一种基于局部决策融合机制的传感器网络故障诊断方法。通过研究局部节点状态信息与网络故障的关联性，采用高效的决策融合模型来检测和推理网络中出现的异常和导致故障的原因。课题完成了基于粗糙集理论的节点故障特征属性提取与约简，建立了与故障类型相对应的决策表；完成了节点诊断证据融合模型研究；提出了局部决策融合的故障诊断算法。不仅提高了故障诊断正确率，也降低了故障诊断开销，并且在针对变电站设备状态监测的应用背景下，诊断模型与算法得到验证。本课题在三年研究期内，共发表论文16篇，其中SCI/EI检索8篇，提交发明专利申请1项，参加IEEE GLOBECOM等国内外学术会议2次。