TBM目前已集成上百个传感器和信息采集系统，可实时记录反映TBM施工状况及工作性能的运行数据，为开展数据驱动研究及实现TBM智能施工提供了良好的契机。在分析近年来TBM机器学习相关研究成果的基础上，综述TBM智能施工的研究进展。首先，以引松工程3标段为例，介绍TBM运行参数及运行数据的基本情况，分析TBM控制参数及掘进循环的各工作阶段划分； 然后，系统归纳TBM掘进过程隧道（洞）围岩智能感知、TBM掘进过程地质灾害预测预警、TBM掘进控制参数辅助决策3方面的研究进展； 最后，结合当前的技术水平和研究现状，探讨目前研究中存在的瓶颈，并提出对后续研究展望的建议，即多源数据互补信息的利用、专业机制知识-数据的融合驱动、新工程/工况应用场景的迁移学习是该领域有待进一步研究的方向。
Tunnel boring machines
(TBMs) are equipped with hundreds of sensors and information acquisition
systems, enabling real-time
recording of operational data that reflects construction status and working
performance. This provides a valuable opportunity for data-driven research and the advancement of intelligent TBM construction.
The authors review recent research achievements in TBM data-driven machine learning and summarize key developments in TBM
intelligent construction as follows: (1) Using TBM 3 bid section of the Songhua
river water diversion project as a case study, the operational parameters and
data of the TBM are introduced. TBM control parameters and the segmentation of
each tunneling cycle stage are analyzed. (2) Recent advancements in three key aspects
of intelligent TBM construction are systematically summarized, including
intelligent perception of tunnel surrounding rock, prediction and early warning
of geological hazards, and assisted decision-making of
TBM tunneling parameters. (3) Based on the current technological landscape and
research progress, existing research bottlenecks are discussed and future
research directions are proposed, including: utilization of multi-source data for complementary information extraction, integration of
domain knowledge with data-driven approaches, and
application of transfer learning across different engineering projects and
working conditions.
