Analysis and Prediction of Deformation of Shield Tunnel Under the Influence of Random Damages Based on Deep Learning DOI Creative Commons

Xiaokai Niu,

Yong Pan,

Wei Li

и другие.

Buildings, Год журнала: 2025, Номер 15(10), С. 1590 - 1590

Опубликована: Май 8, 2025

Shield tunnels in operation are often affected by complex geological conditions, environmental factors, and structural aging, leading to cumulative damage the segments and, consequently, increased deformation that compromises safety. To investigate behavior of tunnel linings under random this study integrates finite element numerical simulation with deep learning techniques analyze predict shield segments. First, a refined three-dimensional model was established, modeling method developed simulate evolution different ratios. Additionally, statistical analysis conducted assess uncertainty caused damage. Furthermore, introduces convolutional neural network (CNN) surrogate enable rapid prediction conditions. The results indicate as ratio increases, both mean its variability progressively rise, instability, demonstrating effect on segment deformation. Moreover, 1D-CNN trained using computation results, predictions test dataset showed excellent agreement FEM calculations. achieved correlation coefficient (R2) exceeding 0.95 an RMSE below 0.016 mm, confirming ability accurately across best our knowledge, finite-element–deep-learning hybrid approach proposed provides valuable theoretical foundation for predicting in-service assessing safety, offering scientific guidance safety evaluation repair strategies.

Язык: Английский

Analysis and Prediction of Deformation of Shield Tunnel Under the Influence of Random Damages Based on Deep Learning DOI Creative Commons

Xiaokai Niu,

Yong Pan,

Wei Li

и другие.

Buildings, Год журнала: 2025, Номер 15(10), С. 1590 - 1590

Опубликована: Май 8, 2025

Shield tunnels in operation are often affected by complex geological conditions, environmental factors, and structural aging, leading to cumulative damage the segments and, consequently, increased deformation that compromises safety. To investigate behavior of tunnel linings under random this study integrates finite element numerical simulation with deep learning techniques analyze predict shield segments. First, a refined three-dimensional model was established, modeling method developed simulate evolution different ratios. Additionally, statistical analysis conducted assess uncertainty caused damage. Furthermore, introduces convolutional neural network (CNN) surrogate enable rapid prediction conditions. The results indicate as ratio increases, both mean its variability progressively rise, instability, demonstrating effect on segment deformation. Moreover, 1D-CNN trained using computation results, predictions test dataset showed excellent agreement FEM calculations. achieved correlation coefficient (R2) exceeding 0.95 an RMSE below 0.016 mm, confirming ability accurately across best our knowledge, finite-element–deep-learning hybrid approach proposed provides valuable theoretical foundation for predicting in-service assessing safety, offering scientific guidance safety evaluation repair strategies.

Язык: Английский

Процитировано

0