基于机器学习的盾构施工地表沉降预测方法研究  

Research on Surface Settlement Prediction Method of Shield Construction Based on Machine Learning

作　　者：孙森 葛爱迪 康建国 庞宇哲 周亚东 Sun Sen;Ge Aidi;Kang Jianguo;Pang Yuzhe;Zhou Yadong(The Eighth Engineering Co.,Ltd.,of The First Highway Engineering Bureau of CCCC,Tianjin 300170,China;College of Civil Engineering,Tianjin Chengjian University,Tianjin 300384,China)

机构地区：[1]中交一公局第八工程有限公司,天津300170 [2]天津城建大学土木工程学院,天津300384

出　　处：《市政技术》2024年第8期142-149,共8页Journal of Municipal Technology

基　　金：国家自然科学基金(51608351);天津市自然科学基金(18JCZDJC10010)。

摘　　要：对长短期记忆网络(LSTM)模型进行了网络结构优化改进,依托几何参数、地质参数和掘进参数等多源数据进行了特征提取,深入分析了隧道盾构施工引发的地表沉降,并对比分析了BP神经网络模型和LSTM模型的预测精度。分析结果表明:LSTM模型相较于BP神经网络模型具有更好的预测能力,与实际工程监测数据更加吻合;在施工过程中,可利用模型预测数据对地表沉降变形提供超前预警,通过调整盾构掘进参数来实现地表变形控制。相关研究结论可为类似盾构施工地表沉降预测提供参考。The network structure of the long-term and short-term memory(LSTM)model is optimized and improved.The feature is extracted by multi-source data such as geometric parameters,geological parameters and driving parameters.An in-depth analysis is conducted on the surface settlement caused by tunnel shield construction.The prediction accuracy of BP neural network model and LSTM model was compared and analyzed.The analysis results show that the LSTM model has better prediction ability than the BP neural network model,and is more consistent with the actual engineering monitoring data;During the construction process,surface settlement&deformation can be pre-warned by the model prediction data.The surface deformation control can be realized by adjusting the shield tunneling parameters.The relevant research conclusions can provide reference for the prediction of surface settlement in similar shield construction.

关 键 词：盾构施工 地表沉降 机器学习 神经网络 预测方法 

分 类 号：U455.43[建筑科学—桥梁与隧道工程] U456.3[交通运输工程—道路与铁道工程]