欢迎光临《计算力学学报》官方网站！

Reliability analysis method of slope stability considering spatial variability of parameters and its application

DOI：

 作者 单位 邮编 易平* 大连理工大学 116024 刘福海 国网经济技术研究院有限公司 李志轩 大连理工大学 王洪志 中国建筑西南设计研究院有限公司山东分院

边坡稳定可靠性分析的难度不仅体现在所涉及的功能函数为高度非线性的隐式函数，还在于土性参数通常具有空间变异性，应采用隐含众多随机变量的随机场模型，导致计算量急剧增加，造成维度灾难。针对这些困难，本文提出基于维度距离策略的主动学习Kriging可靠性分析方法，通过维度距离约束避免样本点彼此邻近而造成信息冗余，提高计算效率；多个数值算例表明，在高维问题中该方法的效率优势尤其明显。将改进的主动学习Kriging方法与物质点强度折减法相结合，建立起一种基于随机场的边坡稳定可靠性分析方法，通过对一个两层边坡的基于平稳随机场和非平稳随机场的稳定可靠性分析，表明所提方法有效减少了计算量，同时也表明考虑土性参数随深度变化的非平稳随机场边坡的失效概率要明显高于平稳随机场边坡的失效概率。

The difficulty of slope stability reliability analysis is reflected in both the fact that the performance function involved is highly nonlinear implicit function, and that soil parameters are usually spatially variable and random field models essentially with many random variables should be adopted, which leads to a sharp increase in computational load and dimensional disaster. Aiming at these difficulties, this paper proposes an active learning Kriging reliability analysis method based on dimensional distance strategy, which avoids the phenomenon of sample points being too close to one another and information redundancy and improves the computational efficiency. Several numerical examples show that the method’s efficiency is more evident in high-dimensional problems. Slope stability reliability analysis based on random field model can then be performed by combining this improved active learning Kriging method and the material point strength reduction method. The stability reliability analysis of a two-layer slope based on stationary and non-stationary random fields shows that the proposed method can reduce the computation cost effectively. It is also shown that the failure probability of the non-stationary random field slope considering the variation of soil parameters with depth is significantly higher than that of the stationary random field slope.
查看/发表评论  下载PDF阅读器