基于D-S证据理论的白水河滑坡状态跃迁研究

发布时间：2018-03-01 07:41

本文关键词： D-S证据理论状态跃迁降雨特征向量　出处：《中国地质大学》2016年硕士论文　论文类型：学位论文

【摘要】：滑坡是危害最大的地质灾害之一,它对公共设施以及人民生命财产安全造成极大的威胁和损失。顺层滑坡是三峡库区最常见的滑坡类型之一。而且该类滑坡一旦失稳,对库区安全造成极大的破坏。研究表明,顺层滑坡处于缓慢蠕滑阶段时,滑坡启动的可能性非常小；而滑坡处于加速变形阶段时,滑坡启动的可能性就大大增加了。因此研究滑坡跃迁到加速变形阶段的概率对滑坡的预测预报有重大的应用价值。本文以白水河滑坡"ZG93"监测点为研究对象,对滑坡状态跃迁到加速变形阶段的概率问题展开了深入的研究。主要的研究内容如下：(1)扩充样本数据。运用工程类比法和人工降雨实验,扩充样本作为先验数据。(2)对先验数据进行预处理。对滑坡位移累计数据、降雨数据、库水位数据分别进行分析处理,对滑坡状态进行划分,提取降雨特征向量。(3)运用D-S证据理论对先验数据进行融合。结合白水河滑坡的地质和水文条件,确定“滑坡是否跃迁到加速变形阶段”的评价指标、识别框架；采用统计数据方法对先验数据进行计算得到各评价指标的基本概率分布,并联合构成基本可信度；利用Dempster合成法则计算基本可信度最终的跃迁概率。本文的主要创新点如下：(1)运用D-S证据理论客观地探究滑坡的状态跃迁问题,不仅保留了专家知识的优点,还将人工智能和数学的客观有效性引进来,能够对滑坡状态跃迁的研究提供一种参考方法。(2)通过对降雨数据的特征分析、特征提取,结合白水河滑坡的地质特征,在保留降雨数据内在联系及规律的基础上,获得了加权求和后的蒸发量、入渗量、径流量的表达式[α β,γ],以此作为每个月的三维降雨特征向量。通过以上研究,在以白水河滑坡为工程实例的研究中发现,运用统计的方法从大量的先验数据中计算得到各个评价指标的基本概率分布。在前期滑坡状态处于等速变形阶段时的跃迁概率最大；在降雨特征向量属于第一类时的跃迁概率最大；在库水位变化属于第一类时的跃迁概率最大。运用D-S证据理论对3个评价指标进行融合,当某月3个评价指标各自的跃迁概率均处在最大时,滑坡状态跃迁到加速变形阶段的概率很大。
[Abstract]:Landslide is one of the most dangerous geological hazards, which poses a great threat and loss to the safety of public facilities and people's lives and property. The bedding landslide is one of the most common landslide types in the three Gorges Reservoir area, and once the landslide is unstable, The research shows that when the bedding landslide is in the slow creep stage, the probability of the landslide starting is very small, while the landslide is in the accelerated deformation stage, The probability of landslide initiation is greatly increased. Therefore, the study of the probability of landslide transition to accelerated deformation stage has great application value for landslide prediction and prediction. In this paper, the "ZG93" monitoring point of Baishui River landslide is taken as the research object. In this paper, the probabilistic problem of landslide transition to accelerated deformation is studied in depth. The main research contents are as follows: 1) expand the sample data, use engineering analogy method and artificial rainfall experiment, The expanded sample is used as prior data to preprocess the prior data. The accumulated data of landslide displacement, rainfall data and reservoir water level data are analyzed and processed, respectively, and the landslide state is divided. Extraction of rainfall characteristic vector. (3) using D-S evidence theory to fuse the prior data. Combining with the geological and hydrological conditions of the Baishui River landslide, the evaluation index of "whether the landslide moves to the accelerated deformation stage" is determined and the identification framework is established. The basic probability distribution of each evaluation index is obtained by using statistical data method to calculate the prior data, and the basic reliability is constructed jointly. The main innovation of this paper is as follows: (1) using D-S evidence theory to explore the state transition problem of landslide objectively, not only the advantage of expert knowledge is retained, but also the advantage of expert knowledge is preserved. The objective validity of artificial intelligence and mathematics is also introduced, which can provide a reference method for the study of landslide state transition. Through the characteristic analysis and feature extraction of rainfall data, combined with the geological characteristics of Baishui River landslide, On the basis of preserving the internal relation and regularity of rainfall data, the expressions [伪尾, 纬] of evaporation, infiltration and runoff after weighted summation are obtained, which are taken as the three dimensional rainfall characteristic vectors of each month. In the study of Baishuihe landslide as an engineering example, it is found that the statistical method is used to calculate the basic probability distribution of each evaluation index from a large number of prior data, and the transition probability is the largest when the landslide is in the stage of isokinetic deformation. The transition probability is the largest when the rainfall characteristic vector belongs to the first category, and the transition probability is the greatest when the water level of the reservoir belongs to the first class. The D-S evidence theory is used to fuse the three evaluation indexes. When the transition probabilities of each of the three evaluation indexes are at the maximum, the probability of transition from landslide to accelerated deformation is very large.
【学位授予单位】：中国地质大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P642.22

【相似文献】