基于BP模型的基坑开挖引起邻近隧道变形预测  

Deformation Prediction of Adjacent Tunnel Induced by Foundation Pit Excavation Based on BP Network Model

作　　者：王皓晨 李飒[1] 林澜 翟超[2] 邢卫民[2] Wang Haochen;Li Sa;Lin Lan;Zhai Chao;Xing Weimin(School of Civil Engineering,Tianjin University,Tianjin 300072,P.R.China;Tianjing Institute of Geotechnical Investigation Survey,Tianjing 300191,P.R.China)

机构地区：[1]天津大学建筑工程学院,天津300072 [2]天津市勘察院,天津300191

出　　处：《地下空间与工程学报》2021年第S02期832-839,847,共9页Chinese Journal of Underground Space and Engineering

基　　金：国家自然科学基金(51478313)

摘　　要：由于邻近地铁隧道的基坑开挖会影响原隧道结构的安全与稳定,故有必要针对基坑开挖过程中邻近隧道的变形情况进行研究。以天津某邻近地铁隧道的基坑工程为背景,提出了一种优化的BP神经网络方法对该问题进行研究。依据基坑及隧道的变形实测数据,分析了影响隧道水平位移的主要因素,将基坑与隧道的相对位置、围护结构的深层水平位移、支撑轴力以及开挖深度作为神经网络输入层的神经元,并确定了隐含层节点数以及有效监测范围。根据上述分析建立了一种可以预测隧道结构水平位移的BP神经网络。通过已训练好的BP网络对二期工程影响下的隧道水平变形进行预测且引入均方根误差RMSE进行分析,得出预测得到的隧道水平位移曲线形态与实测形态大致相同,从而验证了优化后的BP神经网络的可靠性。Because the excavation of foundation pit will affect the safety and stability of adjacent subway tunnel,it is necessary to investigate the deformation of the tunnel during the excavation.Based on afoundation pit engineering near a subway tunnel in Tianjin,an optimized BP neural network is proposed to study this issue.According to the measured deformation data of retaining structure of foundation pit and tunnel,the main factors affecting the horizontal displacement of tunnel are analyzed.The relative position of foundation pit and tunnel,the deep horizontal displacement of retaining structure,support axial force and excavation depth are regarded as the neurons of neural network input layer,and the number of hidden layer nodes and effective monitoring range are determined.According to the above analysis,a BP neural network which can predict the horizontal displacement of tunnel structure is established.Through the trained BP network,the horizontal deformation of the tunnel under the influence of phase II project is predicted,and the root mean square error RMSE is introduced to evaluate the prediction results.It could be found that the predicted horizontal displacement curve shape of the tunnel is roughly the same as that of the measured,which verifies the reliability of the optimized BP neural network proposed.

关 键 词：BP神经网络 输入层因素分析 监测范围 隐含层节点数 

分 类 号：TU753[建筑科学—建筑技术科学]