【作者】 李远耀；

【导师】 殷坤龙；

【作者基本信息】 中国地质大学 ， 岩土工程， 2010， 博士

【摘要】 国内外水库工程特别是大型水库工程中,水库蓄水及库水位周期性的循环涨落,经常会诱发库岸新老滑坡的产生或复活,这类受库水影响、由库岸斜坡孕育的滑坡通常称为库岸滑坡(或水库型滑坡)。大量调查和研究表明,在丘陵和山区水库中,因蓄水运行而造成库岸滑坡产生和复活的现象非常普遍,而且其失稳破坏的后果常常十分严重,有时甚至是灾难性的。目前,在水库库区地质灾害防治工作实践中,开展滑坡的监测预报工作已经成为避免和减轻库岸滑坡危害的重要手段之一在当前的滑坡研究领域,滑坡预测预报也是最重要和最前沿的研究热点之一。尽管经过数十年的发展,滑坡预测预报方法和理论得到了长足发展,在实践方面也取得了一些成功经验,但是大量滑坡预测预报的实践表明,现行的滑坡预测预报理论、模型或判据仍不能对滑坡的变形演化行为和具体的失稳时间做出真正准确的预测预报。原因在于：一方面大多数的预测预报理论、模型或判据还缺乏坚实的地质、力学方面的物理基础,并与个体滑坡的特殊变形演化机制结合不足；另一方面,过于侧重于探索滑坡具体失稳时间的预测模型或判据,往往忽略滑坡在环境因素(如降雨和库水位涨落)变化下变形动态的预测、基于实际监测信息的中长期变形趋势的预测以及滑坡变形演化阶段判别等内容,这种预报思路割裂了滑坡的孕育-变形-失稳破坏的内在演化规律。具体到库岸滑坡的预测预报,在复杂的库水涨落作用下,特别是与降雨等因素叠加后,库岸滑坡的变形破坏机制及其表现形式更是千差万别,难以把握。因此,针对这种复杂性和多样性,仅仅依赖于滑坡具体失稳时间的预测模型或判据,对于库岸滑坡的预测预报而言是远远不够的。鉴于以上问题,论文以三峡库区库首段近80处大型库岸滑坡的长时期位移监测资料和大量野外调查工作为基础,综合采用地质学、滑坡学、岩土力学、可靠度分析理论、数值模拟技术、统计分析方法以及非线性理论,针对三峡库区渐进式库岸滑坡的预测预报问题开展了较为系统的探讨：紧紧围绕库岸滑坡渐进破坏的力学作用过程及其表现形式,结合降雨及库水位变动的联合作用,进行了库岸滑坡渐进破坏概率预测的研究；结合滑坡的实际位移监测信息,进行了滑坡变形演化阶段判别方法和定量失稳预报模型的研究。通过以上一系列的分析和研究,论文主要取得了以下一些成果和结论：(1)基于滑坡分类理论,将滑坡的变形过程和受力状态特征结合,建立了反映滑坡变形时空演化特征的地质模型分类,把滑坡分为渐进推移式滑坡、渐进牵引式滑坡、渐进平移式滑坡、突变推移式滑坡、突变牵引式滑坡和突变平移式滑坡等六种类型。(2)以大量实例滑坡的长时期实际监测资料和大量野外工作为基础,探讨了库岸滑坡渐进破坏的影响因素：发现库首段渐进式滑坡的坡体多由松散堆积物构成,堆积物坡体结构松散、物质成分分布不均,可为滑坡的局部不均匀变形提供变形发展空间,渐进式滑坡的滑带或潜在滑带以碎石土、含碎石粘土或粉质粘土、粘土为主,抗剪强度具有一定程度的峰残强度差；滑坡的坡体结构特别是潜在滑动面的形态与滑坡渐进破坏形式具有很好的对应关系,牵引式滑坡的滑面类型主要为近平直型、下陡-中缓-上陡型和上陡下缓型,而推移式滑坡则主要为上陡下缓型；通过对实例滑坡位移加速过程与降雨和库水位的变化过程的相关分析,指出降雨和库水位变动是区内滑坡渐进破坏的主要动因。并基于土体的渐进破坏理论,总结了降雨和库水位联合作用下库岸滑坡的渐进破坏机制。(3)针对滑坡在降雨和库水位变动条件下的“事件”预报问题,发现传统的滑坡转移概率分析方法,存在相邻两条块间沿不同路径转移概率相等和不能分析非均质滑带土滑坡两个问题；在假设应变软化岩土材料为理想脆性材料和不考虑条间力随机性的基础上,推导了考虑滑带土应变软化特征的滑坡渐进破坏概率分析方法,并结合小概率的极值降雨过程及三峡库区实际库水位变动双重因素联合作用对滑坡渗流场的影响,通过非稳定渗流分析方法,建立了库水位变动及降雨条件下库岸滑坡的渐进破坏概率预测模型。(4)采用极值Ⅰ型分布函数,对库首段巴东站26年来的降雨资料进行了分析,得出其特征性极值降雨的拟合参数和多日累积最大降雨的降雨量与降雨过程特征；运用K-S有限比较法,对2938组抗剪强度参数试验数据分类进行了拟合优度检验,发现三峡库区五类易滑坡地层中滑坡滑带土抗剪强度参数的概率统计特征为：一般均能接受正态分布和对数正态分布,但不同类型滑带土参数指标的最优分布有所不同；总体而言,内摩擦角φ值相对于内聚力c值,最优分布服从正态分布概型的更多,服从对数正态分布的更少；对于参数变异性,无论在何种状态下,内聚力c值的变异系数均大于相应内摩擦角φ值。(5)对实例枣子树坪滑坡降雨和库水位变动条件下渐进破坏概率的预测结果表明：不同的库水位状态,滑坡出现初始破坏的发育部位不同,破坏可能性的大小也不相同,当初始破坏出现于不同降雨条件下时,滑坡渐进破坏的潜在规模不同；若滑坡发生破坏,不同条件下其潜在破坏模式不同,可能推移式破坏,也可能渐进牵引式破坏,但绝大多数情况下不会一次完成渐进破坏过程,不同变形破坏方式下存在不同的阻滑地段或锁固地段；滑坡出现不同形式的渐进性破坏,当以整体破坏概率阀值作为预报判据时,其潜在的危险性是完全不同的,其中,175m库水位叠加50年重现期降雨时潜在危险性最高。通过实例滑坡探讨了的渐进破坏概率预测结果在滑坡监测预报中的运用：与相关判据阀值结合,可开展降雨和库水位变动条件下的滑坡“事件”预报,预报目标为滑坡失稳可能性和潜在的初始破坏位置、变形破坏模式及最危险的破坏规模；并可以指导监测预报中关键性(或控制性)监测点的布设。(6)基于前人总结的滑坡断裂构造生成次序规律和滑坡裂缝分期配套特征,通过滑坡地表裂缝成因机制的分析,构建和完善了滑坡变形演化阶段的地表裂缝判据。基于渐进式滑坡破坏过程的一般特性,将位移速率接近度与滑坡累积加速度和累积加加速度判据相结合,并考虑滑体不同空间位置监测点加速度和加加速度的相互关系,建立了改进的位移动力学参数判据；该判据不仅保留了原判据中对滑坡变形破坏过程在演化时间上规律性的反映,而且考虑了在滑坡不同变形阶段,滑坡体空间变形的发育规律,在实际监测预报中通过与外界影响因素的相关性分析,不仅能对滑坡整体进入加速阶段进行准确的判别,也可以对滑坡局部位移出现“阶跃式”变化的间歇加速过程进行较为准确的判断。(7)提出了滑坡定量失稳预报中的三个基本问题,即：滑坡初始状态的问题、反映滑坡动态信息源的问题和滑坡预测预报模型适用性的问题。指出：整体加速变形阶段的出现是滑坡变形演化过程的重要转折点,滑坡加速变形阶段之前的预测预报均可为中长期预测预报,加速变形阶段之后则开展短临预报；中长期预报可考虑采用处理复杂性问题较为有效的非线性科学的理论和方法建立预测预报模型或判据,以滑坡的变形发展趋势或具体的位移值为主要预报目标(失稳时间可为次要目标)；短临预报可考虑建立确定性较强的模型或判据进行预报,并以失稳时间作为最主要的预报目标,同时,临滑预报必须针对不同滑坡类型选择适宜的模型和方法、并以确切的加速阶段(或临滑加速阶段)的位移数据为基础,才可能得出相对准确的预报结果。(8)从渐进式滑坡变形演化的阶段性出发,提出了中长期变形趋势预测的R/S分析方法、位移预测的时间序列-神经网络模型以及临滑预报的改进Verhulst模型。得出相关结论为：滑坡地表位移监测时间序列集中地反映了滑坡变形趋势的非线性特征,描述和刻画复杂非线性时序的R/S分析方法能有效对其进行分析,但增量位移是更有效的分形结构,它的Hurst指数H值不仅能判断滑坡不同监测点位移在时间上的趋势性及其强度,还能从空间上判断(在同一时刻)滑坡不同部位的变形趋势特征,并进一步提出了判断滑坡加速变形阶段的概念模型判据。位移预测的时间序列-神经网络模型通过对监测位移总量的分解预测,并考虑影响因素与位移变化的响应关系,实现了滑坡变形影响因素与位移动态的综合分析,提高了预测结果的精度和可靠性,位移预测结果通常可与常规的经验性判据阀值结合,进行滑坡失稳预报。将位移速率最大值作为失稳时间预报值时,原始Verhulst预报模型存在一定的滞后误差,提出可将位移速率作为预报参数,位移加速度及位移加加速率分别达到最大值时刻作为失稳时间预报的上下限值,在不明确滑坡启动类型的前提下按上下限时间范围值进行临滑预报更为合理,当监测周期一定时,上下限值的时间差距只与模型参数a值相关。(9)任何单一的预报模型或方法都有其自身的局限性,开展滑坡的综合信息预测预报是实现滑坡相对准确预报的必然途径之一。论文以三峡库区在蓄水以来出现过重大险情的秭归白水河滑坡和卧沙溪滑坡为例,在对它们的地质背景、监测手段、宏观变形历史及累积位移曲线特征进行系统分析的基础上,运用文中提出的方法和模型对它们进行了综合信息预报的实际应用和检验验证分析,结果表明本文提出的基于滑坡渐进破坏过程的相关预测预报方法科学合理,具有一定的实用性和普适性。以上研究成果和结论,不仅为提高和完善滑坡预测预报理论,提供了一些有意义的新思路和新方法,同时对于三峡库区地质灾害监测预警和生态环境建设等实际工作,也具有十分积极的现实意义。更多还原

【Abstract】 Reservoir at home and abroad, especially in the large reservoir, the cycle of periodic fluctuations of water level often induce the production of new or revived old landslide. Such landslides affected by reservoir water and conceived by bank slope are often referred to as bank landslides. A large number of surveys and studies show that the landslides caused by reservoir operation and the phenomenon of generation and resurrection are widespread in the hills and mountains. The consequences of failure often cause very serious damages, sometimes are even catastrophic. At present in the work of geological disaster prevention, landslide monitoring and forecasting is one of the important means to avoid and reduce the bank landslides hazards.In the current research of landslide, landslide prediction is one of the most important and cutting-edge research focuses. Despite several decades of development, the methods and theory of landslide prediction has made considerable progress, and in practice also had some success. However, many landslide forecasting and warning practice has shown that the present theory, model or criterion of landslide prediction still can not make accuracy forecast of the deformation evolution behavior and the specific failure real time. For the reason, on the one hand the most of the prediction theories, models or criterion are lack of solid geological, mechanical and physical basis which is inadequately combined with the specific deformation and evolutionary mechanisms of the single landslide; on the other hand it emphasis too much on exploring the models or criterion to predict the failure time, while ignoring the dynamical prediction in facts of environment (rainfall and water level change) and the prediction of long-term deformation trends based on real time monitoring information as well as landslide evolution stage discrimination. Such forecast ideas separate the inherent evolution law of preparation-deformation-instability. For the prediction of reservoir landslide, under the complex fluctuation of reservoir water, in particular, overlapped with the rainfall and other factors, the deformation mechanism and its manifestations are different. Therefore, in view of the complexity and diversity, it is not enough only depending on the forecasting models or criterion for predicting specific failure time in terms of the prediction of bank landslides.In view of the above problems, based on the long displacement monitoring data of about 80 large-scale bank landslides in head region of the Three Gorges Reservoir Area and a large number of field investigations, integrated with geology, landslide science, rock and soil mechanics, reliability theory, numerical simulation technology, statistical analysis and nonlinear theory, combining the prediction issues of progressive landslide in the Three Gorges Reservoir Area, the paper develops its research systematically:Centering on the mechanical process of progressive failure of reservoir landslide and its manifestations, combined with the joint action effects of rainfall and changes of the reservoir water level, progressive failure probability of reservoir landslide is studied; Besides, according to the displacement information of actual landslide, the method of landslide deformation evolutionary stage identification and quantification prediction model of landslide failure are also studied. The main results and conclusions are as follows:(1) On the basis of landslide classification theory, according to the deformation and stress states of landslide, the geological model classification which can reflect the temporal evolution characteristics of landslide deformation is put forward. The landslide can be grouped into six types, respectively, progressive lapsed landslide, progressive and tractive landslide, progressive translational landslide, mutational lapsed landslide, mutational tractive landslide, mutational and translational landslide.(2) study the reservoir landslide influencing factors based on a large number of actual monitoring data of the landslide instances in long time and a lot of field work:I find that the progressive landslide’s slope body, whose structure is loose and material composition is uneven which can provide deformation development space for landslide local inhomogeneous deformation, is mainly consist of loose deposits in reservoir head. At the same time, I can discovery that the slip zone or potential slip zone of progressive landslides is composed of gravel soil,stony clay or gravel soil and clay mainly whose shear strength has a strength difference between peak intensity and residual intensity in some degree. Moreover, I acquire that landslide slope body especially potential slip surface feature has a good relationship with the form of progressive failure about landslides, the main type of the slip surface in pull landslides is nearly straight, below steep-middle slow-up steep and up steep-bellow slow, but the main type of the slip surface in push landslides is up steep-bellow slow. Finally, according to the correlation analysis to the acceleration process of landslide instances,rainfall and the changing of reservoir water level,I conclude that the combined effect of rainfall and changed reservoir water level is major motivation for the progressive failure of landslides.Meanwhile, based on soil progressive failure theory, I summarize the progressive failure mechanism of landslides under the combined effect of rainfall and changed reservoir water level.(3) Based on the landslide "incident" forecasting under the condition of rainfall and changed reservoir water level,I found the traditional landslide transition probability analysis methods have two problems that the two adjacent blocks sliding along different paths have the same transition probability and don’t analyze the landslides with non-homogeneous slip soil. At the same time, I have derived landslide progressive failure probability analysis method considering slip soil strain softening characteristics on the condition that strain softening geotechnical materials are ideal brittle materials and don’t consider the randomness among blocks. At last, combined with the combined effect of double factors to the influence for landslide seepage field between a very small probability extreme rainfall and the real changed reservoir water level in Three Gorges Reservoir Area, I have established the progressive destruction probability prediction model of bank landslides under the condition including rainfall and the changed reservoir water level according to the unsteady seepage analysis method.(4) Using extreme value I Distribution function the 26 years of rainfall data of Badong in the reservoir head were analyzed, and obtained the fitting parameters of the characteristic extreme rainfall and largest of many days cumulative precipitation and characteristic of rainfall process. Based on the 2938 group of shear strength test data using K-S fit contrast method set classification of goodness of fit test, and found that in the 5 categories of landslide prone strata of Three Gorges reservoir, the probability statistics features of shear strength parameters of slip soil is:generally acceptable normal distribution and lognormal distribution, but the optimal distribution of different types slip soil parameters are different.Generally, internal friction angle (φ) of optimal distribution submit to normal distribution is more than the cohesion (c) value, and lognormal distribution less. For parameter variability, no matter in what state, coefficient of variation of cohesion (c) values were higher than the corresponding value of the friction angle(φ).(5) Take the Zaozishuping Landslide for example, in the conditions of rainfall and changes of reservoir water level, the prediction results showed that:in the state of different reservoir water level, the initial damage occurred in the different parts of the landslides, and the damage possibility was also not the same. When the initial damage occurred in different rainfall conditions, the potential scale of landslide progressive failure was different. If the landslide occurred, the potential damage under different conditions had different modes, may lapse damage or progressive traction damage. But in the vast majority of cases will not be a progressive failure to complete the process. In the different pattern of deformation and failure, there were different anti-skid lots or locking lots. The landslide had many forms of progressive failure. When made the threshold value of whole failure probability as the prediction criterion, the potential risk is completely different, which has highest risk when in the 175m water level and stack the 50-year return period rainfall. We can use the prediction result of potential failure probability in the landslide monitoring and predicting. Combined with the relevant threshold value of criterion, we can make landslide event prediction in the rainfall and changes of reservoir water level. The goals of prediction include the possible, potential initial damage location, failure modes and the scale of the destruction of the most dangerous, and can guide the layout of key point in the monitoring prediction.(6) Based on the generation order rules of landslide fault structure and stage matching characteristics of landslides fissure summarized by previous, this paper build and improve the ground crack criterion on the stage of the landslide deformation and evolution by the analysis of landslide ground crack genetic mechanism. On the basis of general characteristics of progressive landslide failure process, this paper combined the approach degree of displacement rate and cumulative acceleration and cumulative jerk criterion of landslide, and considered the relationship of monitoring point acceleration and jerk on the different spatial location of sliding mass, and built improved displacement dynamics parameters criterion; the criterion is not only retains the original criterion reflect on regularity in the evolution of time of landslide deformation and failure process but also considered development law of landslide spatial deformation in the landslide different deformation stages. In the actual monitoring and prediction, based on the correlation analysis with external factors, it not only accurately discriminant that the landslide slide into the acceleration phase as a whole, but also more accurately discriminant the process of intermittent acceleration when the local displacement of the landslide appeared a "step-type" speed up change.(7) This paper presented three basic problems about quantitative instability prediction of landslide, namely:problems about the initial state of landslide reflect the dynamic information sources of landslide and suitability of landslide prediction model. It pointed out:the emergence of a whole accelerated deformation phase of is an important turning point of landslide deformation and evolution process. Forecast before the stage of landslide accelerated deformation all are the medium and long term forecast, and then carried out short-term prediction after the stage of accelerated deformation. Long-term forecast could considered to established prediction model or criteria using theories and methods of non-linear scientific which is more effective to treat complexity problem, and took the deformation trend of the landslide or concrete displacement value as the main forecast objective(unstable time as the secondary target). Short temporary forecast can consider establishing a stronger deterministic model or criteria to predict and take the forecast failure time as the main target, on the same time, forecasting impending slide must select suitable models and methods according to different type of landslide, and on the basis of the exact displacement data on the speed stage (or before sliding accelerated phase), it may come relatively accurate prediction results.(8) Based on the stage of progressive landslide deformation, it has proposed R/S analysis method to predict long-term deformation trends, displacement time series prediction-Neural Network Model and improved Verhulst Model for temporary sliding forecast. The relevant results are:Landslide time series of surface displacement monitoring has consistently reflected the Nonlinear characteristics of landslide deformation trend, depicted that Complex nonlinear time series of R/S analysis method can be used to do analysis effectively, but the more analysis structure is Incremental Displacement. The Hurst index H can not only decide timely trends and strength of displacement of different monitoring points, but also can judge the space (at the same time) landslide characteristics in different parts of deformation trend.What is more, it furthered the accelerate deformation stage to judge the concept of landslide model criterion. Displacement time series prediction-neural network model forecasted by monitoring the displacement of the total decomposition, and considered the influence factors and displacement response relationship to realize the landslide deformation and displacement of dynamic integrated analysis, so as to improve the prediction accuracy and reliability. Displacement and conventional forecast results usually combine empirical criterion threshold, unstable landslide prediction. Taking the maximum displacement rate as a failure time prediction value, the original Verhulst model has a lagged forecast error, the displacement rate can be raised as a forecast parameter, displacement, filling and emptying rates of acceleration and displacement as the maximum moment of failure time prediction upper and lower limits. When the Startup type is not clear,the temporary landslide forecast according to the value of the minimum and maximum time frame is more reasonable. When the monitoring period is constant, the time gap between the upper and lower limits is only associated with the model parameters a.(9)Any single forecasting model or method has its own limitations, carrying out comprehensive information landslide forecast is a inevitable way to achieve a relatively accurate prediction of the landslide. Taking Wo Shaxi and Zigui Baishuihe landslides both of which appeared significant danger since the Three Gorges reservoir impoundment for example, This thesis analyzed the practical application of comprehensive information forecast and test verification proposed in this thesis based on the systematic analysis of geological background, monitoring tools, macro-deformation history and the accumulated displacement curves. The results show that the proposed process of progressive failure of landslides based on relevant scientific and reasonable prediction method proposed in this thesis has certain practical and universal.The above research results and conclusions can not only enhance and improve the predictive and forecasting theory, provide some interesting new ideas and methods, but also has a very positive significance in monitoring of geological disasters in Three Gorges Reservoir Area for early warning and ecological environment construction and other practical work.更多还原