基于可靠度指标的路基边坡有效随机分析

Effective Random Analysis of Soil Subgrade Slopes Based on Reliability Index

路基边坡的稳定性情况与土体参数空间分布特征有着紧密的关联性,为高效分析不同土体空间变异水平对路基边坡可靠性的影响,提出了一种新的基于集成学习(XGBoost)的机器学习方法作为随机场有限元(RF-FEM)模型的代理模型,通过使用少量但又足够的边坡随机场样本进行适当的训练,可以实现取代计算时间及计算成本上有着严格苛刻要求的蒙特卡罗模拟随机场有限元分析。结果表明,本研究所提出的框架能够准确并且高效地得到路基边坡的失效概率,所提出的XGBoost代理模型方法预测的边坡失效概率(P_f)与直接蒙特卡罗模拟的RF-FEM结果控制在7%以内,但计算成本却显著减低。另外,在路基边坡可靠性分析中,土体抗剪强度参数不同变异水平对路基边坡失效概率影响显著,不能忽视土体参数变异水平的空间变异性。

The characteristics of the spatially distribution of soil strength are directly related to the stability of subgrade slopes.A new method of the XGBoost model is provided as a surrogate model for the random field-finite element(RF-FEM)model to effectively examine the impacts of various soil spatial variation levels on the reliability of subgrade slope.By using a small but sufficient number of random field samples for proper training,it can replace random field-finite element analysis with strict and stringent calculation costs.The results show that the proposed framework can obtain the failure probability of the subgrade slope accurately and efficiently.The failure probability error between the proposed XGBoost surrogate model and the RF-FEM of direct Monte Carlo simulation are controlled within 7%,but the calculation cost is significantly reduced.Furthermore,the different variation levels of soil shear strength parameters have a significant impact on the failure probability of subgrade slope in the reliability analysis,and the spatial variability of soil parameter variation levels cannot be ignored.