Connecting EPBM Data to Ground Movement Data using Machine Learning
Keywords:TBM, Tunneling, Machine Learning, Random Forests
This paper presents a method to estimate tunneling-induced ground movements by connecting the earth pressure balance tunnel boring machine (EPBM) operation data to the ground monitoring data. The proposed method requires no prior assumptions, such as the ground loss and the geologic parameters. This study was conducted using a data set from the State Route 99 (SR99) tunnel project in Seattle, WA. The prediction models were developed using (i) ordinary least squares (OLS) as a parametric linear regression method and (ii) random forests (RF) as a nonparametric nonlinear machine learning method. Segmentation and feature importance analyses were carried out to investigate the influence of EPBM features on the induced ground movements in different ground-machine interaction mechanisms. This study shows that various tunneling-induced ground responses can be estimated solely based on the EPBM feature data and the tunnel spatial geometries. The segmentation and feature importance analyses reveal that each ground response segment has different governing parameters. Features related to the steering and pressure controls appear to influence the induced ground movements during the EPBM passing strongly. These features are not typically considered in conventional tunneling-induced ground movement prediction methods.
Copyright (c) 2023 Dayu Apoji, Zhangwei Ning, Kenichi Soga
This work is licensed under a Creative Commons Attribution 4.0 International License.