基于PLAXIS-BP神经网络的暗挖隧道地表沉降预测  

Prediction of Surface Settlement in Underground Tunnels Based on PLAXIS-BP Neural Network

作　　者：张文旭 李程 陈辉 邵浩 魏东洋 ZHANG Wenxu;LI Cheng;CHEN Hui;SHAO Hao;WEI Dongyang(School of Civil Engineering,Anhui Jianzhu University,Hefei 230601,China;Huainan Traffic Engineering Quality Supervision Station,Huainan Anhui 232002,China;Nanling Transportation Comprehensive Management Service Center,Wuhu Anhui 241300,China)

机构地区：[1]安徽建筑大学土木工程学院,安徽合肥230601 [2]淮南市交通工程质量监督站,安徽淮南232002 [3]南陵县交通运输综合管理服务中心,安徽芜湖241300

出　　处：《兰州工业学院学报》2024年第6期8-13,41,共7页Journal of Lanzhou Institute of Technology

基　　金：安徽省住房城乡建设科学技术计划项目(2021-YT-21)。

摘　　要：为研究浅埋暗挖隧道施工过程中多因素联合作用对地表沉降的影响,准确评估施工风险,基于离心模型试验实测数据,将有限元软件PLAXIS^(3D)与BP神经网络学习算法相结合,建立了饱和黏土地层隧道暗挖施工地表沉降预测模型。通过调节BP神经网络中的隐含层层数和节点数,得到最优神经网络结构,通过增加验证集和敏感性分析进行二次验证,并进行了特征重要性分析,量化各因素对最大地表沉降的影响程度。结果表明:通过调节BP神经网络超参数,所建模型误差小于5%,满足工程精度要求,且预测沉降变化趋势符合工程实际。To study the impact of multiple factors on surface settlement during the construction process of shallow buried and underground excavated tunnels,and accurately evaluate construction risks.Based on the measured data from centrifugal model experiments,a prediction model for surface settlement during tunnel excavation in saturated clay layers is established by combining the finite element software PLAXIS ^(3D) with the BP neural network learning algorithm.By adjusting the number of hidden layers and nodes in the BP neural network,the optimal neural network structure is obtained,and secondary validation is conducted by increasing the validation set and sensitivity analysis.Then,feature importance analysis is performed to quantify the impact of each factor on the maximum surface subsidence.The research results indicate that by adjusting the hyperparameters of the BP neural network,the error of the constructed model is less than 5%,which meets the engineering accuracy requirements,and the predicted settlement trend is in line with the actual engineering situation.

关 键 词：数值模拟 浅埋暗挖 BP神经网络 地表沉降 

分 类 号：TU45[建筑科学—岩土工程]