Deep learning technologies for shield tunneling: Challenges and opportunities
Cheng Zhou,Yuyue Gao,Elton J. Chen,Lieyun Ding,Wenbo Qin
DOI: https://doi.org/10.1016/j.autcon.2023.104982
IF: 10.3
2023-06-28
Automation in Construction
Abstract:Shield tunneling has been prevalent in tunnel construction since its introduction into the field. To take advantage of the massive data generated during tunneling and to assist in engineers' judgement, deep learning models have been widely applied. A comprehensive survey is presented in this paper to organize emerging research and propose future directions. Typical types of data in shield and the corresponding pre-processing approaches are summarized and listed. Specific application scenarios are defined, including the recognition, predication, and control of external environments, shield efficiency, and shield safety. The explainability and generalizability of the applied deep learning models are also analyzed to evaluate their performance. The research challenges are proposed, including the lack of high-quality datasets, limited evaluation of the generalizability of the models, and their limited interpretability. Federated learning, model-based deep learning, and semi-supervised learning are then recommended as potential solutions to these challenges in future research.
construction & building technology,engineering, civil