【作者】 缪新颖；

【导师】 褚金奎；

【作者基本信息】 大连理工大学 ， 微机电工程， 2013， 博士

【摘要】 目前的大坝安全监测系统主要采用“有线”方式,具有采集信号准确、抗干扰性好、产品系列化的特点,但利用有线传感器组成的监测网络布线量大、维护费用高,甚至在一些结构中无法实现布线。无线传感器网络(Wireless Sensor Network, WSN)具有微型化、集成化、节省安装时间和维护费用等优点,可以弥补上述不足。为此本文将WSN应用到大坝安全监测系统当中,并对大坝监测环境中的传感器及节点部署和数据融合等关键技术进行了研究。论文的主要内容包括：(1)以满足大坝采集信号传输距离远、功耗低的需求为目标,构建一种基于分簇无线传感器网络的大坝安全监测系统(Wireless Sensor Network for Dam Safety, DS-WSN)。该系统采用分簇和多跳的网络结构,保证了系统运行的高可靠性和低功耗,系统级ZigBee模块JN5139的应用使传感器节点具有体积小、通信距离远和功耗低的特点。同时,该系统通过汇聚节点可以与互联网和GPRS网络进行连接,以方便数据无线远程传输。与典型无线传感器网络结构比较结果表明,DS-WSN可靠性更高、功耗更低。(2)以大坝监测特点为依据,研究关键断面的确定方法和基于图论理论的传感器及传感器节点两级部署策略。首先采用有限元分析解决关键断面的确定问题,以满足大坝环境节点部署少而精的要求。然后以覆盖效率为目标,研究基于三圆全覆盖理论的大坝关键断面传感器覆盖策略,并利用空洞边界条件,基于最小覆盖圆理论,对覆盖空洞进行修复,以保证大坝断面传感器全覆盖。最后以信号衰减和屏蔽为约束条件,在监测廊道结构图的基础上,提出基于最大通信距离的连通点集生成骨干网的算法,并深入研究骨干网备用节点的冗余部署,可以有效解决大坝环境中传感器节点连通性问题。(3)在DS-WSN基础上,以大坝监测数据需求为目标,以无线传感器节点处理能力为约束条件,提出大坝同质+异质分级数据融合机制。簇成员和簇首节点的同质融合采用简单的阈值判断机制和加权融合算法有效减少数据传输量;与汇聚节点相连的PC机或计算机管理中心的异质融合采用基于主成分分析法的进化神经网络对大坝安全进行预测预报,该模型与传统神经网络预测模型相比,预测精度更高,运行时间更短。(4)开发了DS-WSN监测管理系统,可以实现对各种大坝监测数据的分析和管理,并对DS-WSN进行了监测实验。结果表明DS-WSN组网能力比较强、信号连通性好；所采集的数据传输可靠、精度很高；节点寿命较长,能够满足大坝安全监测需求。更多还原

【Abstract】 Currently, the wired acquisition is mainly used in the dam monitoring system, and it has the characteristics of accurate signals, good anti-interference, and series product. Otherwise, the wired sensor monitoring network has some shortcoming:large wiring, high maintenance costs, and inability of wiring in some specified structures. In this paper, wireless sensor network is used to the dam safety monitoring based on its advantage of the miniaturization, integration, little installation time and low maintenance costs, and key technologies of sensors as well as nodes deployment and data fusion in wireless sensor networks for dam safety are studied. The main contents of the paper include:(1) In order to guarantee the long transmission distance of dam signals and low power consumption, Wireless Sensor Network for Dam Safety (DS-WSN) is presented making full use of the advantages of clustering and multi-hop. The clustering and multi-hop network structure are used in the system, which ensures high reliability and low power consumption. Besides, the JN5139ZigBee modules are applied, which guarantee that the sensor nodes have the characteristics of small size, long transmission distance and low power consumption. Furthermore, the system can connect to the Internet and the GPRS network through the sink node, which facilitates the remote transmission of the wireless signals. DS-WSN has higher reliability and lower power consumption than the typical wireless sensor network architecture.(2) According to the dam monitoring characterizes, the key sections are determined, and the deployment of sensors and sensor nodes is studied based on the graph theory. In order to ensure that fewer nodes can measure effectively, the finite element analysis method is adopted to determine the key sections. To ensure high cover efficiency, coverage strategies of the sensors on the dam key sections based on the full coverage theory of the three circles are researched, and based on the minimum coverage circle theory, the coverage holes are repaired by take advantage of the empty boundary conditions, which ensures full coverage of the key sections. In order to guarantee the effective transmission, on the basis of consideration of the influencing factors such as signal attenuation and shielding, according to the monitoring corridor topology diagram, a backbone network with connected set is raised based on the maximum communication distance. Besides, the redundancy deployment of the backbone network is considered to improve the sensor nodes connectivity in the dam environment. (3) On the basis of DS-WSN, to get proper monitoring data, the data fusion is researched according to data processing capacity of the dam sensors. The data fusion is divided into homogenous fusion and asynchronous fusion. Homogeneous fusion occurs mainly in the cluster member nodes and the cluster head, and a threshold determination mechanism and the adaptive weighted fusion algorithm are applied to reduce the amount of data transmission. While asynchronous fusion mainly occurs in computer management center or PC linked on the sink node, and an evolutionary neural network based on principal components analysis is used to forecast the dam safety. The comparison between the prediction model and the traditional neural networks indicates that the model predicts accurate, and is time-saving.(4) A dam safety monitoring and management system is developed to analyze and manage all kinds of the monitoring data. The monitoring experiment of DS-WSN shows that DS-WSN can network normally, the data transmission is reliable, the collected data are precise, the lives of the WSN nodes are long, which can meet the requirement of the dam safety monitoring.更多还原