基于有限勘察数据的边坡稳定可靠度分析与失稳机制研究

Geological strata and geotechnical properties are inherently spatially variable because of complex geological, environmental, and chemical processes that the strata have undergone. Slope stability analysis with spatially varied geotechnical properties and strata would yield uncertain stability results accordingly. Both stratigraphic uncertainty and geotechnical parameter spatial variability should be properly considered in a slope stability model. The two types of uncertainties, however, cannot be easily quantified because of limited site investigation data in geotechnical engineering practice, which thereby leads to an unreasonable slope stability analysis. This project aims at addressing the difficulty in simultaneously quantifying the two uncertainties from limited site investigation data and investigating the effects of these uncertainties on slope reliability and failure mechanism. Bayesian compressive sampling approach is firstly used to estimate the spatial autocorrelation structure underlying the spatially varied parameters. Secondly, Kriging is employed with the estimated autocorrelation structure to modify the conventional Bayesian compressive sampling approach to construct the conditional random field models (CRFs) of the strata and geotechnical properties. Thirdly, the CRFs are incorporated into a finite element analysis model of slope stability to obtain the probabilistic stability results and slope failure mechanisms. The proposal is illustrated and validated by both simulated and real-life data, based on which the effects of borehole layout schemes on the above uncertainties and thereby the slope stability results are studied. The research results are expected to provide a theoretical guidance on the quantification of the above uncertainties from limited site investigation data and to know the failure mechanisms of slopes subjected to these uncertainties. Moreover, an optimal site investigation scheme is supposed to be reached to enhance the confidence in slope stability results analyzed from limited site investigation data. The project is of great theoretical and practical importance for slope engineering.

实际边坡中存在着地层不确定性和参数空间变异性，从而导致边坡稳定性具有变异性，因此在边坡稳定性分析中应同时考虑这两种不确定性的影响。同时，由于实际工程中勘察数据有限，地层不确定性和参数空间变异性难以有效量化，因而边坡稳定性亦不能准确确定。项目基于有限勘察数据，提出一种可同时量化地层不确定性和参数空间变异性的条件随机场模型，并采用随机有限元方法研究上述不确定性对边坡稳定可靠度和失稳机制的影响。具体是利用贝叶斯压缩抽样方法确定地层和参数的空间变异结构；采用Kriging方法建立地层和参数的条件随机场；然后将条件随机场与有限元方法结合，建立边坡稳定性分析的随机有限元模型。采用模拟数据和实际数据验证模型的有效性，同时研究不同勘察方案对上述不确定性量化以及边坡稳定可靠度和失稳机制的影响。研究结果对量化实际工程中上述不确定性，揭示边坡失稳机制，设计可降低边坡稳定性不确定性的优化勘察方案具有重要价值。

岩土体参数空间变异性和地层边界不确定对边坡概率稳定性分析具有显著影响。本项目基于少量勘察数据，提出一种可同时量化上述两种不确定性的条件随机场模型，采用随机有限元方法研究上述不确定性对边坡稳定可靠度和失稳机制的影响，并通过模拟数据和实际数据验证了模型的有效性，重点在以下三方面取得重要成果：.（1）基于有限勘察数据的地层边界不确定性量化和模拟.采用贝叶斯压缩采样法应用于有限勘察数据下地层边界不确定性的量化和模拟。以模拟的地层边界数据为例，探讨钻孔数据和位置对地层边界不确定性量化和模拟结果的影响，并以实际工程为背景，研究其对岩土结构物（比如边坡）的影响，探讨可以有效地降低地层边界不确定性的勘察方案。.（2）有限勘察数据下地层边界不确定性和参数空间变异性的量化及其对边坡稳定可靠度和失稳机制的影响.将上述理论应用于有限勘察数据下地层边界不确定性和参数空间变异性的同时量化和模拟。以模拟数据为例，研究不同钻孔数据和不同位置情形下的边坡稳定性和破坏机制，探讨确定边坡滑坡影响区域的方法，建立提高边坡可靠度的优化勘察方案，采用实际边坡工程验证方案的有效性和可行性。.（3）基于Kriging理论的高效边坡概率稳定性分析方法.基于传统Kriging理论，在极限平衡方法框架之内，提出自适应Kriging模型、多重Kriging模型以及基于重要域的主动学习Kriging模型用于边坡的高效概率稳定性分析。上述模型克服了传统Kriging模型基于经验选择自相关函数的不足、解决了传统Kriging模型无法评估破坏风险与面对复杂边坡时计算精度不足的问题，推广了Kriging模型在边坡概率稳定性分析的应用。.本项目研究成果，提出的地层边界不确定性量化方法具有很强的工程背景，可用于构造地层剖面；所提的考虑地层变异性与土体参数空间变异性的边坡可靠度分析方法为边坡勘察方案优化提供技术支持；所提三种改进Kriging模型提高了边坡概率稳定性分析的计算效率，具有推广前景。