【作者】 阮殿旭；

【导师】 张晓光；

【作者基本信息】 中国矿业大学 ， 机械电子工程， 2011， 博士

【摘要】 无线传感器网络(Wireless Sensor Network,WSN)是当今国内外备受关注的、由多学科高度综合交叉而来的新型前沿研究领域。无线传感器网络的基本功能是采集并传输节点在监测区域内的感知信息。构建可靠的网络结构，降低网络数据传输量、提高信息传输效率等是无线传感器网络所面对的最大挑战，因此，拓扑控制、路由协议、网络优化等都是以提高网络性能为目的，并成为无线传感器网络重要研究热点。由于无线传感器网络的技术特点，使其在军事、环境、安全、医疗、工农业生产以及智能家居等领域都有着广阔的应用前景。通过研究发现，无线传感器网络在机械故障诊断方面的应用研究尚属起步阶段。尤其在井下大型复杂机械设备上，通过组建无线传感器监测网络来构建安全监测系统，可以降低设备检查的成本，缩短设备停机检修时间，提高效率，结合基于无线传感器网络的故障诊断方法可以发现早期故障征兆和已经发生的故障特征，避免由机械故障引起的井下安全事故。本文对无线传感器网络拓扑控制、路由协议、网络优化方法进行了研究，提出了新的改进方法；详细分析了设备的故障模型及机械故障诊断的主流方法，进行了基于无线传感器网络的故障特征提取及故障类型分类的研究。本文的主要工作体现在以下几个方面：(1)针对煤矿井下工作面的实际要求与无线传感器监测网络需要监测的设备(采煤机、刮板输送机和液压支架群)及环境(温度、湿度、瓦斯浓度等)。详细分析了网络监测设备——采煤机、刮板输送机和液压支架群，提出了需要监测的重要参数及设备状态，并对需要监测的工作面重要环境参数进行了分析。根据监测的设备和环境，设计了井下无线传感器监测网络的总体架构，分析了网络需求和层次结构，同时对井下特殊环境，建立了无线传感器网络通信信道模型，并进行了分析。(2)为了保证煤矿井下无线传感器监测网络的可靠性和连通性，同时减少构建网络拓扑的能耗和时耗，本文提出一种基于自适应分层遗传PID的拓扑控制算法(SAHGA-PID)，在自适应遗传算法的基础上引入分层搜索策略，将其分为两层搜索，提高算法寻优效果。该算法不仅使用了闭环控制思想，而且引入了自适应分层遗传算法，通过仿真表明：该算法可以很好的减少网络启动能耗。使用基于SAHGA-PID的拓扑控制算法进行井下工作面无线传感器网络拓扑控制，可以更快更好的构建网络拓扑结构，减少网络启动过程中的能耗和时耗，快速达到网络连通性要求。保证了井下工作面网络拓扑构建或者重建的效率，为实时监测和后续路由建立提供了良好基础。(3)针对煤矿井下对数据传输可靠性的要求，分析建立了无线传感器网络能耗模型，并在此基础上，提出了井下基于自适应分簇多类分层路由算法(SACMHROUTING)，综合了分布式分簇与集中分簇机制，在分布式分簇时引入距离因子，使得簇首分布更加均衡，同时，在该算法中引入双簇首冗余机制，当簇内节点过多时，可以均衡网络负载；当一个簇首失效时，可以保证网络可靠性和连通性。并在双簇首的可靠性下，提出了基于节点功率调节构建最小路径生成树的路由发现算法，更快建立最小路径生成树，减少能量消耗。结果表明，本文提出的路由算法可以更均匀的分簇，并控制簇内节点数，提高网络负载的均衡性，延长了网络生命周期，增强了网络可靠性。(4)对煤矿井下无线传感器监测网络覆盖控制和设备区的任务调度这两个优化问题进行了详细分析。对已有的无线传感器网络优化算法进行了研究，考虑到众多优化算法的优缺点，本文提出了基于混合蛙跳蚁群的覆盖优化算法，融合了混合蛙跳和蚁群算法的各自优点，弥补了混合蛙跳算法容易发生早熟和蚁群算法早期信息素不足的缺点。仿真结果表明，通过混合蛙跳算法与蚁群算法的融合，对网络搜索最大破坏路径可以有效地对网络覆盖性能进行优化，并进行节点的调整，提高网络覆盖率。同时针对煤矿井下工作面设备区的狭长特性，建立链式无线监测网络的任务调度模型，利用智能遗传算法进行优化，结果表明，通过该模型的优化，可以使得整个链式网络在完成系统任务时，达到时耗和能耗的最优化。(5)阐述了机械故障诊断技术的发展及诊断方法分类，对采煤机关键部件——滚动轴承进行分析，并建立故障特征模型，利用倒频谱对多边带频率故障信号进行分析。研究了对数-线性比变换，主元分析，支持向量机理论，提出了一种基于对数-线性比的概率主成分分析的特征提取算法对轴承信号进行特征提取，和一种基于支持向量机分类器的故障特征分类算法。测试结果表明，基于对数-线性比的PPCA-SVM分类器的故障诊断算法对采煤机轴承故障的诊断效果很好，可以得到很好特征提取效果和故障特征分类效果。更多还原

【Abstract】 As a new and highly interdisciplinary research field, Wireless sensor network isattracting more and more attention at home and abroad. The basic application of WSNis to gather and transfer the perceptive information of sensor nodes in the monitoredregion. How to build a reliable network infrastructure, reduce network data traffic andimprove the efficiency of information transmission are the biggest challenges inwireless sensor network. Therefore, the methods of topology control, routing protocoland coverage control are aiming at the improvement of network performance, andbecoming important research focus.Since the technical characteristics of wireless sensor network, it has broadapplication prospects in military, environment, security, medical treatment, industrialand agricultural production, intelligent living, etc. The research of WSN be applied tothe machine fault diagnosis is still in initial stage. Especially for large complexmechanical devices underground, it can establish the monitoring system with wirelessmonitoring network. The monitoring system can reduce the downtime and costs ofdevice inspection, improve efficiency, what’s more, it can find the signs of earlyfailure and the fault feature existed, avoid the security accident caused by mechanicalfailure.This thesis given an in-depth study on topology control, routing protocol andcoverage control in WSN, and proposed some new improved methods. The faultmodel and fault diagnosis methods are analyzed in detail, and researched the faultfeature extraction and fault type classification. The chief main work and innovation ofpresent study can be drawn as follows.(1) To the actual demands of workface underground, the devices (shearer,scraper conveyor, and hydraulic support group) and environment (temperature,humidity, gas concentration, etc.) need to be monitored, it analyzed the monitoringequipment-Shearer, Scraper and hydraulic support group, listed the importantparameters and device status, and the important environmental parameters needmonitored. According to monitoring parameters of environment and key devices,designed the general architecture of wireless monitoring network underground,analyzed the network requirements and level structure, and established the wirelesstransmission channel model for the special conditions underground. (2) In order to ensure the reliability and connectivity of wireless sensormonitoring network underground, reduce power consumption and time consumptionin the process of network start-up. This thesis proposed a topology control algorithmbased on PID with self-adaptive hierarchical genetic (SAHGA-PID), introduced thehierarchical search strategy to self-adaptive genetic algorithm, divided into two-layersearch, and improved the optimization result of the algorithm. The algorithm used theclosed-loop control theory, and introduced the self-adaptive hierarchical geneticalgorithm. The result shown that the algorithm can reduce the energy consumption ofnetwork boot. Using the topology control based on SAHGA-PID to establish thenetwork topology underground, it will be faster and better in the boot process, and canreduce the energy consumption and time consumption, meet the requirements ofnetwork connectivity quickly.(3) By establishing the model of wireless transmission energy consumption, andon this basis, this thesis proposed a multiclass hierarchical routing protocol based onself-adaptive clustering, it combined distributed clustering mechanism and centralizedclustering mechanism, introduced distance factor to the distributed clustering, makedclusterhead distributing more balance. At the same time, it introduced the redundancymechanism of double-clusterhead. When there are excessive nodes in some cluster, itcan balance the network load; when there is some clusterhead failure, it can guaranteenetwork reliability and connectivity. For searching the optimization path transmit data,cosidered the energy and time consumption, designed the algorithm of constructingthe Min-path tree based on power control. The result shown that the proposed routingprotocol can clustering more uniform, control the number of nodes in the cluster,improve the network load balance, extend the network life cycle, and enhancenetwork reliability.(4) Coverage control and task scheduling of wireless sensor network is analyzedand attributed to the problem of search the best solution. This thesis did a study ofwireless sensor network coverage optimize algorithm existed, and considered theadvantages and disadvantages, purposed the coverage optimization algorithm basedon hybrid frog-leaping and ant colony. The result shown that, by the fusion of hybridfrog-leaping algorithm and ant colony algorithm, to search the greatest damage pathof the network, it can optimize the coverage performance of network effectively, andadjust the position of the node to improve network coverage. Meanwhile, for thenarrow feature of devices area in the workface underground, this thesis establish the task scheduling model of chain-type wireless monitoring network, and optimized themodel with intelligent genetic algorithm. The result shows that, the time consumptionand energy consumption achieve the optimization when entire chain networkcompleted the task with optimizing the model.(5) Described the development of mechanical fault and the classificationdiagnosis methods. Thorough analyzed the rolling bearing that the critical componentsof shearer, and established the model of fault feature. Then, this thesis analyzed thefault signals with multilateral band frequency by Cepstrum. Through did a study oflog-linear, principal component analysis and support vector machine. A method offault feature extraction based on PPCA with Logarithm-linear Ratios is proposed forextracting bearing fault feature, and a classification method based one-to-one supportvector machine classifier for bearing fault feature. The test result is shown that, thefault diagnosis algorithm based on L-PPCA and improved SVM classifier, can workwell for the bearing fault of shearer, and get very good results of fault featureextraction and classification.更多还原