边坡可靠度分析的高斯过程方法  

Gaussian process method for slope reliability analysis

作　　者：苏国韶[1] 肖义龙[1] 

机构地区：[1]广西大学土木建筑工程学院,广西南宁530004

出　　处：《岩土工程学报》2011年第6期916-920,共5页Chinese Journal of Geotechnical Engineering

基　　金：国家自然科学基金项目(50809017);中国博士后科学基金面上及特别资助项目(20080440812;200902354)

摘　　要：针对传统边坡可靠度分析方法的局限性,将高斯过程机器学习与重要抽样方法相结合,提出了边坡可靠度分析的高斯过程方法。利用极限平衡分析构造少量的学习样本,采用基于统计学习原理的高斯过程模型重构边坡隐式功能函数,实现边坡功能函数及其偏导数的显式表达,并构造合理的迭代方式,在计算过程中不断提升高斯过程模型对失效概率贡献较大区域的重构精度,进而应用重要抽样法计算边坡的失效概率与可靠指标。研究结果表明,该方法是可行的,具有较高的计算精度和效率。Considering the limitation of conventional methods,a new method for slope reliability analysis is proposed combined with the Gaussian process（GP） and importance sampling method（ISM）.A small number of learning samples are built by the limit equilibrium method.The implicit performance function is reconstructed by the GP model based on statistical learning.Thus,the implicit performance and its derivatives in slope stability analysis are approximated by the GP model with explicit formulation.An iterative algorithm is presented to improve constantly the reconstructing precision at the important region,which contributes to the failure probability significantly.Then,the importance sampling method is employed to get reliability results of slope.The test results show that the proposed method is feasible and has high accuracy and efficiency.

关 键 词：边坡 可靠度 重要抽样方法 高斯过程 机器学习 

分 类 号：TU47[建筑科学—结构工程]