摘要
针对路基边坡土体空间变异导致有限宽度失稳的特征,将路基长边坡划分为彼此相邻的独立单元;运用适应高维非线性的MARS算法,基于少量代表性边坡样本,建立单元边坡几何、材料参数与可靠指标β的显性表达式;采用可接受部分单元边坡失稳的K/N(F)系统模型,开展长区段路基边坡系统安全性的概率评价.结果表明:长区段路基边坡沿线路纵向的潜在失稳长度主要受土性水平相关距离影响,路基边坡系统宜按50 m进行单元划分;基于MARS算法的预测方程可实现众多单元边坡β的高效计算,验证误差不超过±10%;路基边坡系统失效的发生概率随单元失稳个数增加而快速降低,研究区段内路基单元长度取50 m时,边坡累积失稳个数占比大于0.149%(即4个)的发生概率不大于3%,为偶然事件.
As the spatial variability nature of soil of earth structure slope leads to a limited failure length,a long slope is divided into a few adjacent sections that are independent of each other.The multivariate adaptive regression splines(MARS)algorithm,with a good performance in addressing high-dimensional nonlinear problems,is applied to derive an explicit expression for the reliability indexβwith slope geometry and the material properties,which are obtained from a few representative slope samples.A K/N(F)system model,which can accept partial instability for unit slope,is used to perform a probability-based system safety evaluation of earth structure slope within a long section.The results show that the potential failure length of a long earth structure slope in the longitudinal direction is mainly influenced by the horizontal scale of fluctuation;50 m in length is appropriate for the segmentation of a long slope.A MARS-based prediction equation for calculatingβallows for an efficient computation of all slope segments with error less than 10%.The occurrence probability of slope system failure of earth structure decreases rapidly with the increase of the number of subsystem failures,the probability of an event where more than 0.149%of total slope segments,i.e.,4,fail is less than 3%,indicating a rare occurrence.
作者
罗强
陈伟航
王腾飞
张文生
李政韬
Luo Qiang;Chen Weihang;Wang Tengfei;Zhang Wensheng;Li Zhengtao(School of Civil Engineering,Southwest Jiaotong University,Chengdu 610031,China;Key Laboratory of High-Speed Railway Engineering of Ministry of Education,Southwest Jiaotong University,Chengdu 610031,China)
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2021年第5期841-850,共10页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目(52078435,51878560)
中国博士后科学基金资助项目(2019M663556).
关键词
长区段路基
边坡稳定
系统可靠性
K/N(F)系统
MARS算法
earth structure within a long section
slope stability
system reliability
K/N(F)system
multivariate adaptive regression splines(MARS)algorithm
