应急物资供应关键技术与平台研究

发布时间：2018-05-23 21:05

本文选题：应急物资供应 + 主成分分析　；参考：《西南交通大学》2015年博士论文

【摘要】：本文所说的应急物资,主要指地震灾害救援过程中所需要的设备和生存物资。应急物资供应问题的核心是合理应用最少的物资最大程度地保障人民在灾害情况下的生命和财产安全。由于救援物资通常数量有限,灾害的发生具有较强的随机性,不可能为所有可能的灾害储备相应的救援物资。在有限的物资条件下,合理地应用和分配救援物资,最大程度减少损失,是应急物资供应研究的主要目标。在灾害发生前制定针对不同灾害的应对计划是保证救援过程正确性的必要手段。为顺利进行救援物资的储备,需要先确定物资的数量、质量和种类等基本信息。将历史受灾强度视为信号序列,利用高斯混合模型对该信号建模,可得到更精确的灾害程度预测,计算灾害程度的最大效益期望可进一步得到应急物资需求估计。由于灾后信息阻塞、通讯中断,受灾人员对灾难的恐惧、逃避,在灾难发生后的短时间内人们很难确切掌握灾害程度和灾害分布情况。通过航拍等手段可以获得灾区在灾前与灾后的图像,利用边缘检测和Hough变换等手段,可以客观地计算灾害发生水平。由于人口流动的原因,灾后人们对灾区的具体受灾人数并不清楚,可以通过逃亡人数来估计实际受灾人数。对灾区逃亡路径图进行简化,然后应用向后传递信息的方法,可以进一步推导出灾区对应急物资需求的分布。救援物资储备过程中,需要考虑物资来源、存储成本、运输成本、救援效益等多方面因素。为支持快速决策,引入主成分分析(PCA)方法对选址因素进行简化,仿真实验表明了PCA方法在选址因素简化过程具备有效性与正确性。为快速到达救援点,应急物资应该存放在统一的储备点中,但各种物资也应该分类存储,以便快速地分发,同时节约储存成本。通过最近邻图聚类,将应急物资进行有效地分类,可以在节约物资存储成本的同时兼顾救援效益,为应急物资储备点选择提供数据参考。传统选址方法主要以运筹学方法为主,建模困难,计算量大。本文引入蚁群算法解决应急物资储备点选择问题,降低了选址的运算复杂度,同时防止出现局部最优的陷阱。为最大程度保障灾区人民生命财产安全,不能等到灾区交通状况调查清楚后再进行救援,救援路径通常都是通过外部观测和历史经验做出决定。在救援运输过程中,各车辆运行是一个随机、并发的过程,引入随机Petri网对应急救援的交通运输过程进行建模可以方便地解决灾区交通状况推测问题。在不考虑资源成本、资源不足等矛盾的情况下,包含时间期望的应急物流路径选择,比时间或距离最短的路径选择方法,更具适应性。灾难发生后,对物资捐赠的讨论是灾后互联网等互动媒体的热点主题之一。可以利用人们对捐赠物资讨论的关注度来收集与预测捐赠物资的种类、数量和分布,本文引入主题检测与跟踪技术来对捐赠物资信息进行收集与预测。将各向异性扩散技术引入词网,在词网中体现相关词语的影响,同时保护主题间的边界。提出有限记忆和被动遗忘机制,利用有限的存储空间对词网进行部分索引。不扫描和处理不活动词语,从而实现物资捐赠主题及其词语的快速检测。
[Abstract]:The emergency materials referred to in this article mainly refer to the equipment and materials needed in the process of earthquake disaster relief. The core of the emergency material supply problem is to use the least material to ensure the people's life and property safety in the disaster situation to the maximum extent. It is not possible to reserve the corresponding relief materials for all possible disasters. Under the limited material conditions, the rational application and distribution of relief materials and the maximum reduction of loss are the main objectives of the study of emergency supplies. In order to carry out the reserve of relief materials smoothly, we must first determine the basic information of the quantity, quality and type of materials. The historical disaster intensity is regarded as a signal sequence, and the Gauss mixed model is used to model the signal, and the prediction of the degree of disaster can be more accurate, and the maximum benefit expectation of the degree of disaster can be further obtained by the emergency. The demand is estimated. Because of the post disaster information blocking, communication interruption, the fear of disaster by the disaster, it is difficult to grasp the degree of disaster and the distribution of the disaster in a short time after the disaster. By means of aerial photography, the image of the disaster area before and after the disaster can be obtained by means of edge detection and Hough transformation. The level of disaster occurrence is calculated objectively. Because of the cause of population flow, people are not clear about the number of victims in the disaster area after the disaster. It can be used to estimate the number of actual victims. In the process of rescue material reserve, it is necessary to consider material sources, storage cost, transportation cost, rescue benefit and many other factors. In order to support fast decision, the principal component analysis (PCA) method is introduced to simplify the location factors. The simulation experiment shows that the PCA method is effective and correct in the process of site selection simplification. At the rescue point, emergency materials should be stored in a unified storage point, but all kinds of materials should be classified in order to quickly distribute and save the cost of storage. Through the nearest neighbor graph clustering, the emergency materials can be effectively classified, and the saving cost of materials can be saved and the rescue benefit is taken into consideration, and the selection of the emergency material reserve points can be taken. In this paper, ant colony algorithm is introduced to solve the problem of the selection of emergency material reserve points, reduce the complexity of the location and prevent the local optimal trap, so as to ensure the safety of life and property of the people in the disaster area to the maximum extent, and can not wait for the disaster. In the process of rescue and transportation, the operation of each vehicle is a random and concurrent process. The introduction of random Petri network to the traffic and transportation process of emergency rescue can easily solve the traffic situation in the disaster area. In the case of the contradiction of resource cost and shortage of resources, the choice of the emergency logistics path, which contains the time expectation, is more adaptable than the shortest path selection method. After the disaster, the discussion of the material donation is one of the hot topics of the interactive media, such as the post disaster Internet. The attention degree of material discussion is used to collect and predict the types, quantity and distribution of donated materials. In this paper, the topic detection and tracking technology is introduced to collect and predict the information of the donated materials. The passive forgetting mechanism uses limited storage space to index the word net partially. It does not scan and deal with inactive words, so as to realize the rapid detection of the theme of material donation and its words.
【学位授予单位】：西南交通大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：U116;X4

【相似文献】