Deep learning characterization of rock conditions based on tunnel boring machine data

作     者：Xu Li Min Yao Ji-dong Yuan Yu-jie Wang Peng-yu Li 

作者机构：Key Laboratory of Urban Underground Engineering of Ministry of EducationBeijing Jiaotong UniversityBeijing 100044China School of Computer and Information TechnologyBeijing Jiaotong UniversityBeijing 100044China China Institute of Water Resources and Hydropower ResearchBeijing 100048China China Railway Engineering Equipment Group Co.Ltd.Zhengzhou 450000China 

出 版 物：《Underground Space》 (地下空间（英文）)

年 卷 期：2023年第12卷第5期

页      面：89-101页

核心收录：

学科分类：081406[工学-桥梁与隧道工程] 08[工学] 0814[工学-土木工程] 082301[工学-道路与铁道工程] 0823[工学-交通运输工程] 

基　　金：supported by the National Key R&D Program of China(Grant No.2022YFE0200400) the Natural Science Foundation of China(Grant No.52025094) In addition,we sincerely give our thanks to the data support from the National Program on Key Basic Research Project(973 Program,Grant No.2015CB058100)of China,China Railway Engineering Equipment Group Corporation the Survey and Design Institute of Water Conservancy of Jilin Province,China 

主　　题：TBM Rock condition perception 2D-CNN Weighted loss function Line model 

摘      要：Rock condition perception based on tunnel boring machine(TBM)data is of great importance for not only ensuring tunnel boring safety but also improving construction *** prediction of TBM boring responses(i.e.,torque and total thrust of the cutterhead)largely determines the reliability of rock condition *** this paper,a new architecture of a two-dimensional convolutional neural network(2D-CNN)with a dual-input strategy is proposed to predict the TBM *** TBM Lot 3 of the Yinsong project in Jilin province,China,is taken as the case study in this *** types of models that follow different learning strategies are compared:one is defined as the point model,which only learns data of the stable phase,and the other is defined as the line model,which learns data from both the loading and stable boring *** line model is further improved by the weighted loss function *** results indicate that the strategy of learning data from both the loading phase and stable boring phase and increasing the weight of samples from the stable phase is shown to be optimal in predicting TBM boring *** terms of learning strategies,the line model can learn the influence of active control parameters on passive response parameters,but the point model *** terms of machine learning algorithms,2D-CNN has the best performance,with R2 values of 0.865 and 0.923 for torque and total thrust,*** proposed line model can overcome the problem that the traditional model failed to learn the influence of control *** a model can provide a solid base for the timely optimization of the control parameters in TBM boring process.