基于随机场-贝叶斯的滑面抗剪强度参数反演

为了合理地表征滑面的抗剪强度参数，针对参数的空间变异性与不确定性问题，以三峡库区白水河滑坡为例，基于随机场、自适应条件抽样与可靠度贝叶斯更新方法，建立滑面抗剪强度参数的反演框架. 采用新的全局敏感性方法PAWN，研究不同位置条块滑面抗剪强度参数对稳定性的贡献度. 基于滑面抗剪强度参数随机场模型的反演结果，开展库水位下降工况下非侵入式随机有限元滑坡渗流变形的模拟，分析更新前、后滑面抗剪强度参数的变化特征及空间变异性对滑坡变形的影响. 结果表明：敏感度大的区域分布在滑坡前缘和后缘道路附近；反演后平稳随机场变为非平稳状态，滑坡前、后缘的位移变异性分别增大和减小.

Back analysis of shear strength parameters of sliding surface by using combination method of random field and Bayes theory

The process of back analysis was constructed based on the random field method, adaptive conditional sampling algorithm and Bayesian updating with structural reliability method by taking Baishuihe landslide as a case in order to characterize the spatial variability and uncertainty of shear strength parameters in sliding zone. A new global sensitivity analysis method named PAWN was adopted to analyze the relative influence of shear strength parameters of slices on the factor of safety. The seepage-deformation simulation with non-intrusive stochastic finite element method was conducted based on the inversion results of the random field model for sliding surface. The variation characteristic and spatial variability before and after update was explored, and its effect on the displacement was analyzed. Results show that the area of large sensitivity index is located on the front and trailing edge of landslide. The non-stationary random field is formed after the inversion. The displacement variability on the trailing edge decreases, while that on the front of landslide increases.