基于改进BP神经网络的大断面隧道地表沉降分析  

Analysis on Surface Settlement in Large Section Tunnels Based on Improved BP Neural Network

作　　者：徐磊 马强强 Xu Lei;Ma Qiangqiang(China Railway 20th Bureau Group Fourth Engineering Co.,Ltd.,Qingdao 266000,China;Qingdao University of Technology,Qingdao 266033,China)

机构地区：[1]中铁二十局集团第四工程有限公司,青岛市266000 [2]青岛理工大学,青岛市266033

出　　处：《北方交通》2024年第5期66-70,共5页Northern Communications

摘　　要：文章针对青岛地铁特殊地质的地表沉降问题,以地铁4号线错埠岭站台施工阶段为研究案例,采用基于Midas的数值模拟与BP神经网络相结合的沉降预测方法,建立地表沉降预测模型。该模型改进BP神经网络模型收敛速度慢、易陷入局部极小点的不足,避免数值模拟误差较大、精度不准、可靠性低等因素对分析结果的影响。采用后检差检验法对模型拟合结果进行检验,得出具有较好拟合能力和泛化能力的BP神经网络模型。应用该模型对地铁隧道施工段进行沉降预测,分析结果与实测数据具有相似沉降趋势。Focusing on the surface settlement problem of the special geological conditions of Qingdao metro,taking the construction stage of Cuobuling platform on metro line 4 as the research case,the settlement prediction model is established by combining the numerical simulation based on Midas with BP neural network.The shortcomings of the BP neural network model are improved by the model,such as slow convergence speed and tendency to fall into local minima,so as to avoid the influence of factors such as large numerical simulation errors,inaccurate accuracy,and low reliability and etc.on the analysis results.The post-test method is used to test the fitting results of the model,and the BP neural network model with good fitting and generalization ability is obtained.The settlement prediction of metro tunnel construction sections is carried out by using this model,and the analysis results have the similar settlement trend to the measured data.

关 键 词：BP神经网络 Midas数值模拟 遗传算法 地表沉降 

分 类 号：U456.3[建筑科学—桥梁与隧道工程]