围岩压力释放率智能优化反分析方法  

An intelligent optimization back analysis method for pressure release rate of surrounding rock

作　　者：张研[1] 苏国韶[2] 燕柳斌[2] 曾召田[1] 

机构地区：[1]桂林理工大学广西矿冶与环境科学实验中心,广西桂林541004 [2]广西大学土木建筑工程学院,广西南宁530004

出　　处：《广西大学学报（自然科学版）》2014年第4期802-808,共7页Journal of Guangxi University（Natural Science Edition）

基　　金：国家自然科学基金资助项目(51369007);广西自然科学基金项目(2013GXNSFBA019233);广西自然科学基金重点项目(2010GXNSFD013011);桂林理工大学科研启动费资助项目(002401003413)

摘　　要：为解决隧洞开挖过程中围岩压力释放率难以获得的难题,融合性能优良的粒子群优化算法与高斯过程机器学习方法,借助FLAC3D数值计算软件,提出隧洞围岩压力释放率求取的粒子群优化—高斯过程—FLAC3D方法。将围岩压力释放率求取转化为以岩体力学参数和围岩压力释放率作为决策变量,以计算与实测位移的差作为目标函数的全局优化反分析问题。针对该反分析问题是极值难以获取的高计算代价全局优化问题,采用粒子群算法寻优,在寻优过程中借助高斯过程机器学习模型不断地总结历史经验,预测包含全局最优解的最有前景方向,通过提高粒子群搜索效率并降低适应度评价次数,进而有效地降低围岩压力释放率求取过程中的数值计算工作量。算例研究结果表明,粒子群优化—高斯过程—FLAC3D方法比粒子群优化方法的计算效率提高了近6倍,最后将其应用到锦屏二级水电站B辅助洞围岩压力释放率反分析,获得该工程围岩压力释放率为0.41,占围岩总压力的比重较大,应给予足够重视。Aiming to solve the problem that the pressure release rate of surrounding rock is difficult to obtain during the excavation, a novel cooperative optimization algorithm based on particle swarm optimization （PSO） and Gaussian process machine learning （GP） is proposed, and the algorithm is combined with the FLAC3D to develop a new method called PSO-GP-FLAC3D for the pressure release rate of surrounding rock. The method transforms the problem to a global optimization back analysis problem that treats the error between geodesic displacement and computed displacement as an object function and the pressure release rate of surrounding rock and mechanical parameters as decision variables. However, the extreme value of this problem is difficult to obtain and the cost is expen-sive. GP is used to predict the most promising solution before searching for the global optimum solu-tion using PSO. Through enhancing the searching efficiency of PSO and reducing the times of fitting evaluation, the cost of calculation during the back analysis of pressure release rate of surrounding rock is remarkably reduced. The results of a numerical example show that the computational efficien-cy of PSO-GP-FLAC3D method is increased by nearly 6 times compared with that of PSO. The pro-posed method is applied to auxiliary tunnel B of Jinping Ⅱ power station in China. The pressure re-lease rate obtained for this project is 0. 41 that accounts for a very high share of total pressure of sur-rounding rock, which should be given enough attention.

关 键 词：围岩压力 释放率 反分析 高斯过程 粒子群 

分 类 号：TU457[建筑科学—岩土工程]