【作者】 汤罗圣；

【导师】 殷坤龙；

【作者基本信息】 中国地质大学 ， 地质工程， 2013， 博士

【摘要】 中国是地质灾害发育最严重的国家之一,主要类型有滑坡、崩塌、泥石流、地面塌陷、地裂缝、地面沉陷等六类。中国地调局统计数据显示,仅在2012年八月,全国共发生地质灾害5752起,其中滑坡4841起、崩塌412起、泥石流415起、地面塌陷56起、地裂缝25起、地面沉降3起；造成人员伤亡的地质灾害26起,共造成53人死亡、25人失踪、75人受伤,直接经济损失32.3亿元。根据国土资源部中国地质调查局官方网站全国地质灾害通报的数据,2004年-2011年期间地质灾害六类地质灾害每年造成直接经济损失超过17亿元,伤亡人数超过400人,且在六类地质灾害中,滑坡灾害发育最为广泛,比例高达50.60%～86.11%,占全年地质灾害一半以上。中国每年因滑坡灾害造成的经济损失就在10亿美元以上,依据中国地调局地质灾害通报数据,2009仅全国特大型和大型滑坡灾害共发生16起,直接经济损失达到1.90亿元人民币；2011年死亡失踪10人以上或直接经济损失1亿元以上的重大灾害发生7起,直接经济损失5.99亿元人民币。滑坡灾害不仅造成巨大的经济损失,还严重危害人民的生命安全,1963年的意大利瓦依昂滑坡造成2600多人的死亡；1983年甘肃省东乡县洒勒山滑坡摧毁四个村,死亡220人,重伤22人；据印度报道2000年西藏易贡藏布高速巨型滑坡共造成印度30人死亡,100多人失踪,50,000人无家可归；2003年湖北秭归千将坪滑坡造成24人死亡,1100多人无家可归；2005年菲律宾南莱特省滑坡灾难造成1800多人伤亡堆积层滑坡是滑坡中分布最为广泛、规模大、暴发频率高、突发性强、持续危害性较大的一类致灾体,该类滑坡通常发生在第四系及近代松散堆积层中,其滑体物质一般由次生堆积体,如崩积物、崩坡积物及冲积与崩坡积混合物堆积而成,滑动面一般为堆积层与下覆基岩的接触面,分布于长江三峡、黄河中上游、及香港、广东、福建等地,以长江三峡发育最广泛。根据《滑坡防治工程勘察规范》(DZ/T0218-2006),堆积层滑坡包括滑坡堆积体滑坡、崩塌堆积体滑坡、崩滑堆积体滑坡、黄土滑坡、粘土滑坡、残坡积层滑坡、人工填土滑坡,对于三峡库区,堆积层滑坡主要发育滑坡堆积体滑坡、崩塌堆积体滑坡、崩滑堆积体滑坡、残坡积层滑坡四类。三峡库区因其独特的地质构造作用及地质环境特征导致滑坡大量发育,拒不完全统计,仅在长江上游地区100万km2范围内,就发育滑坡1736个,其中64%为堆积层滑坡,而在三峡库区二、三期治理和监测的崩塌滑坡灾害点中,堆积层滑坡点约占80%。根据统计分析,这些堆积层滑坡在降雨及三峡库区库水位调度作用下大部分目前处于潜在不稳定状态,在降雨和库水位持续作用下其发展趋势为不稳定状态。对于三峡库区来说,滑坡失稳破坏不仅造成滑坡体上人员财产损失,同时会堵塞航道,更重要的是其次生涌浪灾害造成的危害,意大利瓦依昂滑坡就是最好的证明。因此,三峡库区堆积层滑坡的稳定直接关系到库区人员的生命财产安全和水库的正常运营,三峡库区堆积层滑坡的稳定性与预测预报研究具有重要的理论和实际意义。本论文以三峡库区堆积层滑坡为研究对象,在统计分析三峡库区堆积层滑坡发育规律的基础上,利用室内试验、推广Bayes方法、深部位移监测数据定量分析、数值模拟与位移监测反分析、神经网络模型等方法对滑坡抗剪强度参数的选取和预测进行研究,同时采用强度折减法、Rosenblueth计算方法以及点安全系数法运用FLAC3D数值模拟软件对滑坡的整体与局部稳定性进行了探讨,在此基础上采用滑坡GPS专业监测、滑坡深部位移监测、数值分析模型和非线性模型等研究考虑库水位和降雨等外界因素影响的滑坡位移预测模型、滑坡时间预报模型和判据。通过以上研究,最终得到以下结论：(1)统计分析得到了三峡库区堆积层滑坡的分布规律、发育特征及变形破坏特征。三峡库区堆积层滑坡主要发育在三叠系中统巴东组泥岩和侏罗系中统沙溪庙组泥岩和砂岩互层等易滑地层中,在秭归、巴东、巫山、奉节、云阳、万州六个区县分布最广,主要为特大型和大型滑坡,其中大型滑坡最为发育,且大部分为中深层滑坡,顺向坡最为发育,剖面形态多为凸形,且主要发育在中倾坡(10°～40°)中,其物质组成以残积(Qe1)、坡积(Qdl)、崩积(Qcol)、滑坡堆积(Qdel)的第四系粉质粘土夹碎石为主,滑坡前缘发育在200m高程以下,后缘发育在725m高程以下,三峡库区堆积层滑坡地表累积位移曲线形态主要有稳定型、匀速型、收敛型、加速型、阶跃型和回落型六类,深部位移曲线主要有“V”型、“D”型、“B”型、“r”型、“钟摆”型和“复合”型六类,主要存在蠕滑-拉裂、滑移-隆起-整体下滑、滑移-拉裂-剪断三种变形破坏模式。(2)建立了三峡库区堆积层滑坡抗剪强度参数选取和预测的方法。通过对滑坡抗剪强度参数室内试验、推广bayes方法反分析方法、数值方法与位移监测反分析方法、滑坡深部位移监测数据定量分析基础上的滑坡抗剪强度参数反分析方法、滑坡抗剪强度参数BP神经网络预测方法的研究,并以滑坡实例验证,得到了三峡库区堆积层滑坡抗剪强度参数选取和预测方法的原理及适用条件：①室内试验因取样扰动、试验条件、试验误差以及滑坡空间差异性等原因导致试验结果存在很大的不确定性,其可靠性受到很大影响,且对于三峡库区堆积层滑坡来说,滑坡抗剪参数一般处于峰值强度和残余强度之间、饱和与天然状态并存,试验结果不能直接应用于滑坡稳定性分析等领域,一般作为反分析方法的基础数据。②推广bayes方法反分析滑坡抗剪强度参数的分布形式重点在于先验函数的选取,目前一般以区域统计规律作为先验函数,主要适用于滑坡滑带土室内试验组数较少的滑坡,反分析结果主要用于计算滑坡的破坏概率,对滑坡稳定性进行定量评价。③滑坡抗剪强度参数数值方法与位移监测反分析方法是以滑坡室内试验数据或区域统计规律为基础,以滑坡监测位移及其位移趋势为判断依据,采用数值模拟手段选择滑坡最优抗剪强度参数组合。该方法能够应用于三峡库区大量未进行详细勘察仅进行专业监测的滑坡,但其结果易受数值模拟方法软件缺陷、滑坡专业监测结果准确性的影响。④滑坡深部位移监测数据定量分析基础上的滑坡抗剪强度参数反分析方法将滑坡滑带从完整到完全破坏的过程视为滑坡抗剪强度从峰值强度变为残余强度的过程,并采用内聚力和内摩擦角等比例折减。该方法主要适用于进行深部位移监测且位移监测曲线具有典型曲线特征的滑坡,同时需以室内试验数据或区域统计规律为基础。⑤滑坡抗剪强度参数预测是以区域统计规律建立的抗剪强度参数与滑坡基本物理力学参数之间的关系为基础,采用神经网络模型建立滑坡抗剪强度参数的预测模型。该方法适用于仅进行滑坡基本物理力学试验而未进行滑坡抗剪强度参数试验的滑坡,特别是由于时间、试验设备等原因而未进行残余强度试验的滑坡。(3)以三峡库区典型堆积层滑坡为例,建立了三峡库区堆积层滑坡整体与局部稳定性评价方法。以三峡库区典型堆积层滑坡-重庆市万州区塘角1号滑坡为例,首先通过滑坡现场调查及专业监测成果分析确定滑坡的变形情况,同时建立滑坡三维地质模型。然后采用强度折减法和Rosenblueth计算方法对滑坡三维整体稳定性系数和三维破坏概率进行了研究,同时将其结果与二维稳定性系数和二维破坏概率进行了对比分析,从定性和定量的角度综合分析滑坡的整体稳定性。最后采用数值分析及点安全系数法对滑坡各部位的稳定性系数分布、滑坡位移场、滑坡应力应变场、滑坡塑性区分布进行分析研究,并将其研究结果与现场专业监测和宏观变形调查进行对比验证,以此研究滑坡的局部稳定性。滑坡实例研究结果表明：滑坡三维稳定性计算结果比二维计算结果大,滑坡三维破坏概率比二维破坏概率要小；滑坡计算主剖面不能代表滑坡整体,在滑坡整体稳定性评价时宜采用三维计算方法；滑坡整体稳定性和局部稳定性差别很大,在滑坡稳定性评价中不仅要计算滑坡整体稳定性,而且还要根据现场调查及滑坡专业监测的分析结果对滑坡的局部稳定性进行评价,特别是对于大型堆积层滑坡。(4)从考虑库水位、降雨等外界因素的角度,建立了三峡库区堆积层滑坡位移预测、时间预报模型和方法。以考虑三峡库区堆积层滑坡主要外界影响因素-降雨和库水位为出发点,以滑坡整体与局部稳定性评价结果为依据,以数学力学分析、数值模拟等为手段,以R/S分析方法、加卸载响应比理论等为研究方法,建立了三峡库区堆积层滑坡位移预测、时间预报模型和方法：①以R/S分析方法中赫斯特指数H偏离0.5的程度来确定滑坡位移时间序列中趋势项位移和周期项位移的比例,然后采用神经网络模型和多项式拟合来分别预测滑坡的趋势项位移和周期项位移。该方法充分考虑了降雨和库水位对滑坡位移的影响,适用于三峡库区堆积层滑坡的位移预测。②以相邻时刻的库水位差值为加卸载增量,以对应滑坡累积位移加速度相邻时刻的差值为加卸载响应增量,以加卸载响应比与1的大小对滑坡失稳破坏时间进行预报的滑坡加卸载响应比时间预报模型重点考虑了三峡库区特殊的人类工程活动-库水位调度,对研究三峡库区堆积层滑坡时间预报具有重要意义。③将三峡库区堆积层滑坡简化为单一直线形滑动面滑坡,运用前述滑坡抗剪强度参数与滑坡滑带完整性指标的关系,采用剩余推力法建立滑坡稳定性系数与滑带完整性指标之间的关系,以滑坡稳定性系数等于1时的滑带完整性指标作为滑坡时间预报判据,该方法考虑了滑坡抗剪强度随滑坡变形不断变化的特征,经滑坡实例验证,可作为一种有效的预报方法进行进一步探讨。④采用数值模拟手段,以滑坡有限元计算不收敛、滑坡塑性区贯通及滑坡关键位移监测点位移发生突变为滑坡失稳破坏判据,以滑坡失稳破坏时滑坡关键位移监测点的位移和滑坡极端工况条件下降雨量大小作为滑坡失稳破坏判据的数值分析滑坡时间预报判据研究方法,充分考虑了降雨和库水位对滑坡的影响,为滑坡时间预报判据研究提供了一种新的思路,但由于数值模拟手段中本构模型、滑坡边界条件、滑坡计算参数等与滑坡实际情况存在一定的差别,导致滑坡模拟结果与滑坡实际情况存在一定的出入,所以对于该方法的实用性,还需进行不断探索。更多还原

【Abstract】 China is one of the most serious countries, which develops six main types geological disasters including landslide, collapse, debris flow, ground collapse, ground crack and ground subsidence. Statistics data of China Geological Survey shows that only in August2012, the national total of geological disasters is5752, including4841landslides,412collapse,415debris flow,56ground collapse,25ground crack and3ground subsidence,twenty six of which caused casualties,containing that53people were killed,25people were missed,75people were injured, and direct economic loss was3.23billion yuan. According to the national geological disaster data in the official website of China geological survey, Ministry of Land and Resources, during the period from2004to2011, the main six types geological disaster every year caused a direct economic loss more than＄1.7billion, casualties more than400people, and in the six kinds of geological disaster, landslide was developed most widely, whose proportion was as high as50.60%～86.11%, accounted for more than half of the geological disasters.Annual economic loss caused by landslide disasters in China is more than＄1billion, according to the eological hazard reporting data of China Geological Survey, only in2009,16oversize and large landslide disasters were developed, and the direct economic loss was190million yuan, and in2011the major disasters which caused more than10people killed and missed or the direct economic loss exceed100million yuan were7, and the direct economic loss was599million yuan. Landslide disasters not only cause huge economic losses, but also cause serious damage to people’s life safety, in1963Vajont landslide in Italian killed more than2600people, in1983Sa Leshan landslide in Dongxiang county in Gansu province destroyed four village, killed220people and injured22people, India reported in2000YiGongCangBu high-speed giant landslide in Tibet were killed30people, more than100people were missed and caused50000homeless in India, in2003Qian Jiangping landslide in Zigui country in Hubei province killed24people and caused more than1100homeless, in2005Southern Leyte landslide disaster in Philippines caused casualties more than1800people. The colluvial landslide is a kind of hazard-affected body which is the most widely distributed and large scale developed, whose outbreak frequency is high, abruptness is strong and continuous harm is large. This kind of landslide always happens in the quaternary and modern loose accumulation, whose sliding body material is composed with the secondary deposit, such as colluvial, colluvial deposit, alluvial deposit and colluvial deposit, whose sliding surface is the contact surface between accumulation and bedrock, which distributes in the Yangtze three gorges, the Yellow River shelter-forest, and Hong Kong, Guangdong, Fujian and other places, and the most widely in the three gorges of Yangtze river. In accordance with Landslide control engineering survey specification (DZ/T0218-2006), the colluvial landslide contians landslide deposit landslide, collapse deposit landslide, avalanche deposit landslide, loess landslide, clay landslide, the diluvial layer landslide, and artificial filled soil landslide. For the three gorges reservoir area, the colluvial landslide mainly develops landslide deposit landslide, collapse deposit landslide, avalanche deposit landslide, and the diluvial layer landslide.Because of its unique geological tectonics and geological environment characteristics,a large number of landslides are developed in the Three gorges reservoir area. According to incomplete count, only in the upper Yangtze river area of one million kilometer, there has developed one thousand,seven hundred and thirty six landslides, sixty four of which are colluvial landslides, and in two and three stage control and monitoring landslides disaster or collapses in the Three gorges reservoir area, the rate of colluvial landslides is about80%. According to the statistical analysis, recently most of colluvial landslides lie a potential unstable state under rainfall and reservoir water level scheduled, and they will trend to a unstable state with time. For the three gorges reservoir area, landslide instability and failure not only causes loss of life and property, at the same time it will jam channel, more important is the damage of secondary surge disaster, Vajont landslide in Italy is the best proof. Therefore, the stability of colluvial landslides in the three gorges reservoir directly relates to the people’s life and property safety and normal operation of the reservoir,studying on the stability and prediction of colluvial landslides in the Three gorges reservoir area has an important theoretical and practical significance.This paper takes colluvial landslides in the Three gorges reservoir area as the research object, uses laboratory test, the promotion of Bayes method, deep displacement monitoring data quantitative analysis, numerical simulation and displacement monitoring back analysis, the neural network model to study selecting and predicting landslide shear strength parameters based on the statistical analysis to development law of colluvial landslides in the Three gorges reservoir area. At the same time, adopts the shear strength reduction method, Rosenblueth calculation method and Unit safety method to discuss whole and local stability of landslides by FLAC3D numerical simulation software, then employs the landslide GPS professional monitoring, landslide deep displacement monitoring, numerical analysis model and a nonlinear model to research landslide displacement prediction model, landslide time prediction model and the criterion considering he reservoir level and rainfall. Through the above research, finally get the following conclusion:(1) Get the distribution law, development characteristics and deformation failure characteristics of colluvial landslides in the Three gorges reservoir area by the statistical analysis.Colluvial landslides in the Three gorges reservoir area are developed in easy sliding formation such as mudstones of badong group in the middle Triassic and mudstone and sandstone interbedding of shaximiao group in the middle Jurassic, widely distributed in six districts and counties, zigui, badong, wushan and fengjie, YunYang and wanzhou. Most of the colluvial landslides is the super large, large landslides and deep landslides, whose profile form is convex and composition is silty clay containing gravel of residual deposit(Qel),slope deposit(Qdl), colluvial deposit(Qcol) and landslide accumulation (Qdcl) in Quaternary, and the large-scale landslide, dip slope and middle pour slope (10°～40°) is the most developed. The front and rear edge of colluvial landslides is200m elevation and725m elevation. The landslide surface cumulative displacement tracing patterns have six type, stable type, uniform type, convergence speed type, acceleration type, step type and back type, and deep displacement curves mainly contain "V" type,"D" type,"B" type,"r" type,"pendulum" type and "composite" type including three deformation failure modes,the creep-rip, slip-uplift-the overall downturn, slip-rip-cut.(2) Set up the selection and prediction methods of shear strength parameters for colluvial landslides in the Three gorges reservoir area.Through studying laboratory test, the promotion of Bayes method, numerical simulation and displacement monitoring back analysis method, deep displacement monitoring data quantitative analysis method and the BP neural network prediction method for the landslide shear strength parameters, taking typic colluvial landslides in the Three gorges reservoir area as examples, obtain the principle and applicable condition of selection and prediction methods for colluvial landslide shear strength parameters in the three gorges reservoir area:①as the result of sampling disturbance, test conditions, test error and landslide space diversity or other reasons, laboratory test results has a lot of uncertainty and its reliability has a tremendous influence, and for colluvial landslides in the three gorges reservoir area, landslide shear parameters are generally between the peak strength and residual strength, saturation and natural state coexist, test results cannot be directly applied to the landslide stability analysis etc, generally as a back analysis method based data.②the key point for the promotion of Bayes method back analyzing the distribution form of landslide shear strength parameters lies in the selection of a Prior function which generally take regional statistical rules as, and this method is mainly suitable for the landslides that lack of enough laboratory test group, and the analysis results are used to calculate the landslide failure probability for the landslide stability quantitative evaluation.③the numerical simulation and the displacement monitoring back analysis method for landslide shear strength parameters takes the laboratory test data or regional statistical rule as the foundation and the landslide monitoring displacement and displacement trend as judgment basis, and adopts the numerical simulation method to select the optimal combination of landslide shear strength parameters. This method can be applied to a large number of landslides in the three gorges reservoir area which lack of a detailed survey just have a professional monitoring, but the result is susceptible to be affected by numerical simulation software defect and the accuracy of landslide professional monitoring results.④the landslide shear strength parameter back analysis method based on quantitative analysis to the landslide deep displacement monitoring data,takes the process of landslide slip zone from full to complete destruction as the process of landslide shear strength from the peak strength to residual strength, and uses a proportional reduction for cohesion and internal friction angle. This method is mainly suitable for landslides which have a deep displacement monitoring and whose displacement monitoring curves have a typical curve characteristics, taking laboratory test data or regional statistical rule as the foundation.⑤landslide shear strength parameter prediction is a method applying the neural network model to predict landslide shear strength parameters, based on the relationship between shear strength parameters and landslide basic physical and mechanical parameters by regional statistical rule. This method is only applicable for landslides which have basic physical and mechanical test but lack of shear strength parameters test, especially for the landslides not conducting residual strength tests because of the time, test equipment and other reasons.(3) Establish a whole and local stability evaluation method for colluvial landslides taking typic ones as examples in the Three gorges reservoir area.Taking TangJiao NO.1landslide which is a typic colluvial landslide of three gorges reservoir area in WanZhou district ChongQing province as an example, firstly determine the deformation of the landslide through the landslide site investigation and professional monitoring results analysis. and establish the landslide three-dimensional geological model. Secondly adopt the shear strength reduction method and Rosenblueth calculation method to study the3-d whole stability coefficient and failure probability stability for the landslide, and compare the results with2-d stability coefficient and failure probability results to comprehensively analyze the whole stability for the landslide from the point of qualitative and quantitative view. Finally, apply numerical analysis and unit safety method to analyze and research stability coefficient distribution, displacement field, stress and strain field, and plastic zone distribution of the landslide, and make a comparison validation with field professional monitoring and macroscopic deformation survey to study the local stability of the landslide.Landslide case study results show that landslide3-d stability calculation result is bigger than those of2-d calculation results, but landslide3-d failure probability is smaller than those of2-d calculation results; Landslide calculation main section does not represent the whole landslide, and it is better to use3-d calculation method for landslide whole stability evaluation; Landslide whole stability has a big difference with local stability, the landslide stability evaluation not only has to calculate the integral stability of the landslide, but also assess landslide local stability according to analysis results of the field investigation and landslide professional monitoring, especially for large colluvial landslide.(4) From the view of considering reservoir level, rainfall or other external factors, set up the displacement prediction, time prediction model and method for colluvial landslides in the three gorges reservoir area.Taking considering the mainly external influence factors rainfall and reservoir water level of colluvial landslides in the three gorges reservoir area as a starting point, landslide whole and local stability evaluation results as a judgement, mathematical mechanics analysis and numerical simulation as a means, R/S analysis method, loading and unloading response ratio theory as the research method, establish the displacement prediction, time prediction model and method for colluvial landslides in the three gorges reservoir area:①confirm the proportion of trend displacement and cycle displacement in landslide displacement time series according to the deviation degree between Hurst index contained by R/S analysis method and0.5, then use the neural network model and polynomial fitting to predict landslide trend displacement and cycle displacement separately. This method fully consider the influence of rainfall and reservoir water level,and is suitable for predicting displacement of colluvial landslides in the three gorges reservoir area.②the landslide loading and unloading response ratio time prediction model forecasts landslide instability and failure time based on the relative size between loading and unloading response ratio and1, taking reservoir water level difference between adjacent time for the loading and unloading increment, corresponding landslide cumulative displacement acceleration difference between adjacent time for the loading and unloading response increment. This method fully considers the special human engineering activities in the three gorges reservoir area-reservior level scheduling, which has a great significance for colluvial landslide time prediction in the three gorges reservoir area.③simplify the colluvial landslide in the three gorges reservoir area as a single linear sliding surface landslide, utilize the above-mentioned relationship between landslide shear strength parameters and slip zone integrity index, adopt he residual thrust method to establish the relationship between the landslide stability coefficient and slip zone integrity index, and take slip zone integrity index while the landslide stability coefficient is equal to1as a landslide time prediction criterion. According to landslide examples test, this method considers the changing characteristics of landslide shear strength with landslide deformation and can be further discussed as an effective prediction method.④the landslide time prediction criterion research methods by numerical analysis takes the finite element calculation misconvergence, landslide plastic area cutthrough and displacement mutations of displacement monitoring key point as the landslide instability and failure criterion, and takes the landslide displacement of displacement monitoring key point when landslide is at an instability and failure state, and the rainfall size under extreme working condition as landslide instability and failure criterion, applying the numerical simulation means. This method fully considers the influence of rainfall and reservoir water level for landslides and provides a new idea to study landslide time prediction criterion. However, as a result of the certain difference among constitutive model, landslide boundary conditions, and landslide calculation parameters in numerical simulation means with landslide actual situation, lead to exist a certain access between landslides simulation results and the actual situation. Therefore, it need to discuss continuously for the practicability of this method.更多还原