基于BOA-XGBoost算法的盾构隧道沉降预测研究  

Research on Settlement Prediction of Shield Tunnel Based on BOA-XGBoost Algorithm

作　　者：唐飞 李朋 胡俊杰 葛寨辉 TANG Fei;LI Peng;HU Jun-jie;GE Zhai-hui(Wuhan Metro Bridge and Tunnel Management Co.,Ltd.,Wuhan 430061,China;China Railway Siyuan Group Engineering Operation and Maintenance Co.,Ltd.,Wuhan 430061,China)

机构地区：[1]武汉地铁桥隧管理有限公司,武汉430061 [2]中铁四院集团工程运维有限责任公司,武汉430061

出　　处：《价值工程》2025年第13期122-124,共3页Value Engineering

基　　金：中铁四院科技研发项目(2023JSSY01)。

摘　　要：盾构隧道沉降预测对隧道设计、施工以及运维有着重要意义,通过机器学习算法以及历史监测数据能够有效预测隧道沉降情况。通过贝叶斯优化算法(BOA)、极限梯度提升算法(XGBoost)建立盾构隧道沉降预测模型。利用长期沉降监测数据对所建模型以及对比模型进行了留一法交叉验证,验证结果表明BOA-XGBoost预测模型的预测精度与稳定性均为最佳,R2、RMSE、MAPE分别为0.929、5.346、6.942%,明显优于对照组中的其他预测模型,能够用于运营地铁隧道长期沉降预测。The prediction of shield tunnel settlement is of great significance to tunnel design,construction,operation and maintenance,and can effectively predict tunnel settlement through machine learning algorithms and historical monitoring data.The Bayesian optimization algorithm(BOA)and the limit gradient boosting algorithm(XGBoost)were used to establish the settlement prediction model of shield tunnel.The results show that the prediction accuracy and stability of the BOA-XGBoost prediction model are the best,with R2,RMSE and MAPE of 0.929,5.346 and 6.942%,respectively,which are significantly better than other prediction models in the control group,and can be used for long-term settlement prediction of operating subway tunnels.

关 键 词：盾构隧道 沉降预测 机器学习 贝叶斯优化 极限梯度提升 

分 类 号：U121[交通运输工程]