A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Deep learning technologies for shield tunneling: Challenges and opportunities
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.
Graphical abstract Display Omitted
Highlights Research related to deep learning technologies for shield tunneling is critically reviewed and analyzed qualitatively. Typical types of monitoring data and three main application scenarios in shield tunneling are summarized and listed. Performance of deep learning models and related challenges are analyzed for shield tunneling. Future research opportunities for existing challenges are proposed for deep learning models in shield tunneling.
Deep learning technologies for shield tunneling: Challenges and opportunities
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.
Graphical abstract Display Omitted
Highlights Research related to deep learning technologies for shield tunneling is critically reviewed and analyzed qualitatively. Typical types of monitoring data and three main application scenarios in shield tunneling are summarized and listed. Performance of deep learning models and related challenges are analyzed for shield tunneling. Future research opportunities for existing challenges are proposed for deep learning models in shield tunneling.
Deep learning technologies for shield tunneling: Challenges and opportunities
Zhou, Cheng (author) / Gao, Yuyue (author) / Chen, Elton J. (author) / Ding, Lieyun (author) / Qin, Wenbo (author)
2023-06-10
Article (Journal)
Electronic Resource
English
Shield tunneling technologies in Japan
| British Library Conference Proceedings | 2005
Technologies of Shield Tunneling Along Special Alignment
| Springer Verlag | 2022
A Dynamic Model of Machine Learning and Deep Learning in Shield Tunneling Parameters Prediction
| Springer Verlag | 2023