隧道围岩变形预报的高斯过程机器学习模型  

Deformation Forecasting for Tunnel Rock by Gaussian Process Machine Learning Model

作　　者：苏国韶[1,2] 张研[1,2] 燕柳斌[1,2] 

机构地区：[1]广西大学土木建筑工程学院,南宁530004 [2]广西大学工程防灾与结构安全教育部重点实验室,南宁530004

出　　处：《桂林理工大学学报》2010年第4期551-555,共5页Journal of Guilin University of Technology

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

摘　　要：针对隧道围岩变形过程是一个高度复杂的非线性时间序列的问题,以隧道围岩变形历史资料作为知识源,通过构建围岩变形时间序列,采用高斯过程模型建立隧道围岩当前变形与历史变形的非线性映射关系,在此基础上分别提出了基于静态知识库与基于动态知识库的隧洞围岩变形预报的高斯过程模型。工程实例应用研究表明,围岩变形预报的高斯过程机器学习模型是可行的,无需建立复杂的岩体力学模型,根据历史实测资料就能够对隧道围岩变形做出合理预报,与基于静态知识库的高斯过程模型相比较,基于动态知识库的高斯过程模型的预报精度更高。As deformation of surrounding rock mass is a highly complicated nonlinear time series problem,GP models based on static and dynamic for deformation forecasting are proposed.Based on historical deformation and establishing deformation of surrounding rock mass time series,the nonlinear relationship between present deformation of surrounding rock mass and historical deformation of surrounding rock mass is built by GP model.The case study shows that the GP machine learning model forecasting deformation of surrounding rock mass of tunnel is feasible.Without establishing complicated model of rock mass mechanics,the GP model can forecast nonlinear deformation of surrounding rock mass of tunnel reasonably according the historic measured information of deformation.The forecasting result based on static knowledge base is superior to that based on dynamic knowledge base.

关 键 词：围岩变形 高斯过程 机器学习 预报 

分 类 号：U451.2[建筑科学—桥梁与隧道工程] U456[交通运输工程—道路与铁道工程]