圆弧破坏型岩质边坡安全系数快速估计的高斯过程模型  

Fast Estimation of Safety Factor for Circular Failure Rock Slope Using Gaussian Process Model

作　　者：苏国韶[1,2] 

机构地区：[1]广西大学土木建筑工程学院,广西南宁530004 [2]中国科学院武汉岩土力学研究所岩土力学与工程国家重点实验室,湖北武汉430071

出　　处：《应用基础与工程科学学报》2010年第6期959-966,共8页Journal of Basic Science and Engineering

基　　金：国家自然科学基金(50809017);中国博士后科学基金(20080440812);岩土力学与工程国家重点实验室开放研究基金(Z110601)

摘　　要：高斯过程是新近发展起来的一种新的机器学习方法,对处理复杂非线性问题具有良好的适应性.针对边坡非线性系统的复杂性,为实现边坡安全快速设计和稳定性评价的工程实践要求,在高斯过程回归模型的基础上,提出了一种圆弧破坏型岩质边坡安全系数估计的高斯过程模型.该模型不必建立复杂的力学计算模型,而是利用高斯过程的自学习功能,通过对工程实例先验知识进行学习,建立圆弧破坏型岩质边坡安全系数与其各种影响因素之间的非线性映射关系,然后利用贝叶斯推理规则估计边坡安全系数.工程实例研究的结果表明,该模型是可行的,可以快速准确地给出具有概率意义的圆弧破坏型岩质边坡安全系数.Gaussian process（GP）,which is a newly developed machine learning technology based on statistical theoretical fundamentals and Bayesian theory,has become a power tool for solving highly nonlinear problems.Aiming to that slope engineering is a highly complicated nonlinear system,a new model,namely GP model for estimation of safety factor of circular failure rock slope,is proposed based on GP regression theory in order to meet the requirements of fast design and economical stability evaluation in engineering practice.Without complicated mechanics computation process,through learning the empirical knowledge coming from real engineering,the complicated nonlinear mapping relationship between slope safety factor and its affected factors is established easily using GP model.Then,slope safety factor under novel condition is estimated by Bayesian inference of GP method.The results of case study indicated that the model was feasible and very easy to be implemented.Accurate and probabilistic results of slope safety factor under different conditions can be estimated very fast and economically using the model.

关 键 词：边坡 安全系数 高斯过程 机器学习 

分 类 号：TV22[水利工程—水工结构工程]