最小二乘支持向量机-粒子群算法在地下厂房围岩参数反分析中的应用  

Application of Least Squares Support Vector Machine-Particle Swarm Algorithm to Back Analysis of Surrounding Rock Parameters of Underground Powerhouse

作　　者：杨继华[1] 齐三红[1] 郭卫新[1] 张党立[1] YANG Jihua;QI Sanhong;GUO Weixin;ZHANG Dangli(Yellow River Engineering Consulting Co.,Ltd.,Zhengzhou 450003,Henan,China)

机构地区：[1]黄河勘测规划设计有限公司,河南郑州450003

出　　处：《隧道建设（中英文）》2018年第11期1800-1806,共7页Tunnel Construction

摘　　要：为准确确定地下厂房围岩的弹性模量、泊松比、黏聚力、内摩擦角、侧压力系数等参数,以正交设计、最小二乘支持向量机和粒子群算法等现代数学方法为基本手段,建立基于位移增量的围岩参数反分析方法。以CCS水电站大型地下厂房为研究背景,通过工程地质条件研究选取8#机组剖面作为分析对象,采用二维弹塑性有限元方法建立地质结构分析模型。以地下厂房洞室群分层开挖多点位移计实测位移增量为依据,对CCS水电站地下厂房区域围岩力学特性及地应力场特征进行反分析。研究结果表明:主厂房第Ⅵ层与第Ⅰ层开挖和主变室第4层与第1层开挖所产生的位移增量计算值与多点位移计实测值吻合较好,最大相对误差小于10%,说明采用最小二乘支持向量机和粒子群算法相结合的反分析方法在工程上是可行的,且效果较为显著。In order to accurately determine the surrounding rock parameters of underground powerhouse,i.e.deformation modulus,Poisson′s ratio,cohesion,angle of internal friction and lateral pressure coefficient,the back analysis method for surrounding rock parameters based on displacement increment is established by modern mathematical methods like orthogonal design,least squares support vector machine and particle swarm algorithm.According to engineering geological condition analysis,the cross-section of unit No.8 of Coca-Codo Sinclair(CCS)Hydropower Station is selected as the study object,and the geological structural analysis model is established by 2D elastic-plastic finite element method.The back analysis of mechanical characteristics of surrounding rock and geostress field of CCS Hydropower Station is carried out based on monitored displacement increments of layered excavation of underground powerhouse caverns.The study results show that the computation values of displacement increments caused by the excavation at layerⅥandⅠof main powerhouse and layer 4 and 1 of converter room coincide with that of the monitoring values,and the maximum relative error is less than 10%,which indicates that the back analysis method of least squares support vector machine and particle swarm algorithm is feasible and effective.

关 键 词：地下厂房 最小二乘支持向量机 粒子群算法 有限元模拟 位移增量 反分析法 

分 类 号：U452[建筑科学—桥梁与隧道工程]