An Artificial Neural Network-Based Surrogate Model for Rapid Safety Assessment of the Segment–Ring Beam Connection Under Combined Void and Damage Defects

The segment–ring beam connection in shield tunnels represents a structurally complex and vulnerable zone, where the combined effects of voids behind the lining and localized ring beam damage, under the service environment of groundwater and train-induced vibration, significantly threaten structural safety. This study investigates the mechanical behavior and load-bearing capacity loss of this connection under combined void and damage defects through FLAC3D numerical simulations. A total of 104 calculation cases were established, considering variations in void location, damage location, and damage severity, while the groundwater conditions and train-induced vibration were kept constant as the fixed service environment. Based on the numerical database, a multi-view feature-enhanced mixture-of-experts prediction model, termed SRB-BCNet, is proposed. The model integrates a multi-view feature extractor, an adaptive attention mechanism, and a mixture-of-experts network, with the maximum reduction in the safety factor as the core evaluation metric. The results indicate that the haunch void exerts the most significant effect on structural deformation and stress, with the maximum radial displacement reaching 7.81 mm and the maximum principal stress reaching 3.09 MPa, while ring beam damage locally reduces stiffness and alters the stress field. Under combined effects, the displacement magnitude increases substantially to an extreme value of 12.2 mm, and the minimum safety factor drops to 26.1, indicating a pronounced reduction in structural safety reserve. The SRB-BCNet model demonstrates strong predictive performance, achieving an out-of-fold cross-validation average R2 of 0.9696, RMSE of 1.2219 and MAE of 0.9997, while the in-sample overall R2 reaches 0.9845. Ablation experiments confirm that the MoE network is the most critical component; its removal reduces the R2 to 0.2135. The sensitivity analysis indicates that the degree of ring beam damage and the void location behind the lining are the dominant controlling factors: as the residual strength coefficient increases from 0.5 to 0.9 (i.e., the damage severity decreases), the safety factor reduction decreases by approximately 18.6 percentage points, while the influence of the damage location is relatively limited (a maximum difference of 1.65 percentage points among the four damaged locations). The model also outputs a data-driven uncertainty estimate to support rapid screening of defect configurations, while the systematic calibration of this estimate remains a subject of future work. Overall, this study provides a practical surrogate model for the rapid safety assessment of the segment–ring beam connection for the specific tunnel geometry, soil profile and fixed service environment considered, serving as an efficient alternative to time-consuming numerical simulations.

Authors

Institutions

Publication Details

Journal
Buildings
Published
2026-11-19
DOI
https://doi.org/10.3390/buildings16224624
Primary Topic
Geotechnical Engineering and Underground Structures
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An Artificial Neural Network-Based Surrogate Model for Rapid Safety Assessment of the Segment–Ring Beam Connection Under Combined Void and Damage Defects

Hongchuan Xing, Jing Qin, Libin Wang, Chengping Zhang
Buildings
Geotechnical Engineering and Underground Structures
article

An Artificial Neural Network-Based Surrogate Model for Rapid Safety Assessment of the Segment–Ring Beam Connection Under Combined Void and Damage Defects

Hongchuan Xing, Jing Qin, Libin Wang, Chengping Zhang
article en

Abstract

The segment–ring beam connection in shield tunnels represents a structurally complex and vulnerable zone, where the combined effects of voids behind the lining and localized ring beam damage, under the service environment of groundwater and train-induced vibration, significantly threaten structural safety. This study investigates the mechanical behavior and load-bearing capacity loss of this connection under combined void and damage defects through FLAC3D numerical simulations. A total of 104 calculation cases were established, considering variations in void location, damage location, and damage severity, while the groundwater conditions and train-induced vibration were kept constant as the fixed service environment. Based on the numerical database, a multi-view feature-enhanced mixture-of-experts prediction model, termed SRB-BCNet, is proposed. The model integrates a multi-view feature extractor, an adaptive attention mechanism, and a mixture-of-experts network, with the maximum reduction in the safety factor as the core evaluation metric. The results indicate that the haunch void exerts the most significant effect on structural deformation and stress, with the maximum radial displacement reaching 7.81 mm and the maximum principal stress reaching 3.09 MPa, while ring beam damage locally reduces stiffness and alters the stress field. Under combined effects, the displacement magnitude increases substantially to an extreme value of 12.2 mm, and the minimum safety factor drops to 26.1, indicating a pronounced reduction in structural safety reserve. The SRB-BCNet model demonstrates strong predictive performance, achieving an out-of-fold cross-validation average R2 of 0.9696, RMSE of 1.2219 and MAE of 0.9997, while the in-sample overall R2 reaches 0.9845. Ablation experiments confirm that the MoE network is the most critical component; its removal reduces the R2 to 0.2135. The sensitivity analysis indicates that the degree of ring beam damage and the void location behind the lining are the dominant controlling factors: as the residual strength coefficient increases from 0.5 to 0.9 (i.e., the damage severity decreases), the safety factor reduction decreases by approximately 18.6 percentage points, while the influence of the damage location is relatively limited (a maximum difference of 1.65 percentage points among the four damaged locations). The model also outputs a data-driven uncertainty estimate to support rapid screening of defect configurations, while the systematic calibration of this estimate remains a subject of future work. Overall, this study provides a practical surrogate model for the rapid safety assessment of the segment–ring beam connection for the specific tunnel geometry, soil profile and fixed service environment considered, serving as an efficient alternative to time-consuming numerical simulations.

BuildingsVol. 16(22)
Beijing Jiaotong University (CN), China Institute of Water Resources and Hydropower Research (CN)
Geotechnical Engineering and Underground Structures
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon