雅康高速—泸康段沿线泥石流危险性评价

【作者】 杨洋；

【导师】 杨艳梅；

【作者基本信息】 西南石油大学 ， 地学信息工程， 2017， 硕士

【摘要】 雅康高速是连接川西甘孜藏族自治州与内地的第一条高速公路,该高速公路将成为川西地区的交通枢纽。公路建设在四川盆地西部边缘与青藏高原东部边缘的过渡带,区内地形复杂、山高谷深、坡度较陡,对公路的建设和养护造成了极大的困难。公路穿越了龙门山断裂带、鲜水河断裂带、安宁河断裂带的交汇处,此区域内降水量大,地震活动频繁且地震烈度高,断裂活动极为强烈,人类经济活动增加,这些都为泥石流灾害的发育提供了极为有利的条件,对公路结构形成了巨大的威胁。深入研究雅康高速所在地区泥石流形成机理和发育特征,建立可靠的泥石流危险性评价体系,可以提供灾害评估和预警信息,对公路设施的保护和人民生命财产的安全的保障有着重要的科学意义和实际价值。本文通过对研究公路的实地考察和资料翻阅,探索研究区泥石流灾害的发生机理,确定对研究区泥石流灾害形成贡献较大的评价因子;将流域作为危险性评价单元,并以流域单元为单位提取各评价因子数据;通过遗传算法对BP神经网络进行优化,利用优化后的BP神经网络模拟研究区泥石流流域发生泥石流灾害的概率;在GA-BP神经网络的基础上建立针对研究区的泥石流危险性评价模型,最后实现对研究公路沿线的泥石流危险性评价。本文的主要研究工作如下:(1)以DEM为底图,结合研究区实际情况,利用ArcGIS10.2平台建对研究区进行流域划分。最终将研究区划分成集水面阈值大小分别为1km2、3km2、5km2和10km2的子流域。获取研究区子流域后再根据泥石流的运动特征,公路泥石流成灾形式以及Google Earth的三维真实地形叠加分析三个方面来筛选可能对公路造成影响评价单元。最后,确定的研究公路泥石流危险性评价单元有31个。(2)以研究区泥石流成灾条件为基础,构建研究公路泥石流评价因子体系,分别为:地形起伏度、平均坡度、断层密度、切割密度、地震烈度、年均降水量、地层岩性、年降水变差系数和归一化植被指数(NDVI)。本文将这9个因子以流域为单元进行了数据的提取和分析。(3)本文利用BP神经网络建立了泥石流危险性评价模型,并通过遗传算法对BP神经网络的初始权值和阈值进行优化,以此提高网络的性能。将经过训练的GA-BP神经网络的性能与未优化的BP神经网络进行对比,结果经过优化的BP神经网络的性能要高于未优化的BP神经网络,由此判定该网络模型适用于研究公路的泥石流危险性评价。(4)以划分好的流域单元对研究公路泥石流危险性进行评价,得出有1条流域为极度危险;高度危险的流域有12条;中度危险的流域有6条;轻度危险的流域有12条。更多还原

【Abstract】 Yakang highway which is the first highway connecting Ganzi Tibetan prefecture with the mainland will become the transportation junction in West Sichuan area.It is constructed at the transitional zone between western margin of Sichuan basin and eastern margin of Tibetan plateau.Great difficulties for highway construction and maintenance are caused by the complex terrain,lofty mountains,deep valleys and steep slope in this zone.The highway passes through the intersection of Longmen Mountain fault zone,Xianshuihe fault zone and Anninghe fault zone.Since favorable conditions for debris flow development are created by high precipitation,frequent and intense earthquake,extremely strong faulting and increased human economic activities in this intersection,highway construction as well as local people’s lives and property are threatened.Therefore,deep research on the formation mechanism and development features of debris flow in the highway region,and the establishment of reliable debris flow hazard assessment system can provide disaster evaluation and early warning information.They also have great scientific significance and practical value.Through on-the-spot investigation and document browsing,the mechanism of debris flow is realized.Assessment factors contributing more largely to the debris flow hazard in searching area are determined.Assessment factor database is established in units with drainage basin as hazard assessment unit.31 debris-flow watersheds which may have impact on Yakang highway are selected based on the movement features of debris flow.The occurrence probability of debris flow in debris-flow watersheds is simulated through the establishment of BP neural network functional approximation.Global searching ability of genetic algorithm is used to optimize the initial weight value and threshold value of BP neural network,strengthen the network convergence,and improve the prediction accuracy.Debris flow hazard assessment is modeled for GA-BP neural network in research area.Finally,hazard assessment of debris flow along highway with drainage area as unit is realized,and the assessment result is strongly reliable with a certain practical significance.The main research tasks in the thesis are as follows:(1)DEM as the base map and the real situation in research area as consideration,ArcGIS10.2 platform is used to divide the research area into watersheds.The research area is divided into sub-watersheds with the threshold value of 1km2,3km2,5km2 and 10km2 respectively.After the sub-watersheds in research area are acquired,the assessment units influencing the highway are selected according to the movement features of debris flow,the disaster form of debris flow and Google Earth’s overplay analysis of three-dimensional real terrain.Finally,31 assessment units are determined to study the debris flow hazard along the highway.(2)Based on the condition of debris flow formation in the research area,assessment factor system for highway debris flow is constructed,which includes relief,average gradient,fault density,cutting density,seismic intensity,mean annual precipitation,annual precipitation variation coefficient and NDVI.In this thesis,the 9 factors are used to establish assessment factor database with watershed as a unit.(3)BP neural network is applied to build assessment model of debris flow hazard.Besides,the initial weight and threshold value of BP neural network are optimized with genetic algorithm in order to improve the network performance.Compared with non-optimized BP neural network,the optimized one through training has a better performance.Therefore,it is concluded that this network model is suitable for the study of hazard assessment of debris flow along highway.(4)The divided watershed units are used to assess the debris flow hazard along the research highway.It is concluded that there is 1 extremely dangerous watershed,12 high-dangerous watersheds,6 moderate-dangerous watersheds,and 12 low-dangerous watersheds.更多还原