Scholarlink

The shield machines are extensively employed in the large-scale construction of urban underground space, and the safe application plays a crucial role in the construction process. As underground space construction moves toward deep burial, further research is needed to explore the interactions between more complex soil structures and shield machines. This paper develops a physics-constrained LSTM model for predicting the cutterhead torque of the shield machine. This model introduces the shear strength of soil as prior knowledge and uses the characteristics of the geotechnical-shield tunneling system as constraints to address real-time changes in geotechnical engineering issues. The model is validated on the Beijing East Sixth Ring Road Reconstruction Project undertaken by CCCC First Highway Engineering Co., Ltd. The results show that the prediction accuracy of the model reaches about 85%, which is about 5% higher than the traditional LSTM model.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on AI Prediction Methods for Shield Tunneling Parameters in the East 6th Ring Road Reconstruction Project in Beijing

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yujin Sun,
  • Changyun Yin,
  • Mianfeng Lin,
  • Shuobofang Yang

摘要

The shield machines are extensively employed in the large-scale construction of urban underground space, and the safe application plays a crucial role in the construction process. As underground space construction moves toward deep burial, further research is needed to explore the interactions between more complex soil structures and shield machines. This paper develops a physics-constrained LSTM model for predicting the cutterhead torque of the shield machine. This model introduces the shear strength of soil as prior knowledge and uses the characteristics of the geotechnical-shield tunneling system as constraints to address real-time changes in geotechnical engineering issues. The model is validated on the Beijing East Sixth Ring Road Reconstruction Project undertaken by CCCC First Highway Engineering Co., Ltd. The results show that the prediction accuracy of the model reaches about 85%, which is about 5% higher than the traditional LSTM model.