以张家口市东太平山隧道工程项目为背景,基于MIDAS软件建立有限元模型,利用退火算法优化GA-BP神经网络,设计正交试验并构建GASA-BP神经网络,通过隧道拱顶沉降、水平收敛和仰拱隆起3个参数的模拟值对围岩的弹性模量、黏聚力、内摩擦角进行反演分析.结果表明:使用GASA-BP神经网络反演所得到隧道拱顶沉降、水平收敛和仰拱隆起的模拟值,与实测值的最大相差分别为6.60％、19.80％、2.16％;与BP神经网络相比,GASA-BP神经网络整体反演精度更高,反演所得围岩参数精度处于合理区间,在此参数下建立的有限元模型可以对工程实际进行较好地模拟.
Based on the project of East Taiping Mountain Tunnel in Zhangjiakou City,a finite element model with MIDAS software was established.GA-BP neural network by use of annealing algorithm was optimized,and then orthogonal test and construct GASA-BP neural network to was designed.According to the values of simulations of tunnel arch roof settlement,horizontal convergence and invert arch uplift,the inversion analysis on elastic modulus,cohesive strength and internal friction angle of the surrounding rock were carried out.The results show that,for the value of simulation of tunnel arch roof settlement,horizontal convergence and invert uplift,which is obtained from the inversion by use of GASA-BP neural network,and the value of the actual monitoring,the maximum difference between them are 6.60％,19.80％and 2.16％,respectively.Compared with BP neural network,GASA-BP neural network can provide higher accuracy of inversion,and the precision of surrounding rock parameters obtained by the inversion is within a reasonable range.The finite element model established under these parameters can contribute to good simulation of engineering practice.
张俊武;牛洪军;张朋举;刘小军
中交一公局第一工程有限公司,北京 102205河北省土木工程诊断、改造与抗灾重点实验室,河北张家口 075000||河北省寒冷地区交通基础设施工程技术创新中心,河北张家口 075000||河北建筑工程学院土木工程学院,河北张家口 075000
交通运输
隧道工程;破碎围岩;神经网络;退火算法;反演分析;正交试验;沉降监测;围岩参数
tunnelling;broken rock;neural network;annealing algorithm;inversion analysis;orthogonal test;settlement monitoring;parameter of surrounding rock
《路基工程》 2024 (002)
183-190 / 8
10.13379/j.issn.1003-8825.202306015
评论