Advanced identification method for adverse geological conditions in TBM tunnels based on stacking ensemble algorithm and Bayesian theory

Accurately identifying adverse geology ahead of the tunnel face in real-time is essential for guaranteeing the safe and efficient excavation of tunnel boring machines (TBMs). This study presents a novel real-time identification method for adverse geological conditions ahead of TBM tunnel face based...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Tunnelling and underground space technology Jg. 163; S. 106741
Hauptverfasser: Wu, Zhijun, Huo, Dongbo, Chu, Zhaofei, Liu, Xiqi, Weng, Lei, Xu, Xiangyu, Li, Zheng
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.09.2025
Schlagworte:
Adverse geology prediction
Bayesian theory
Ensemble learning
Sliding double-windows method
TBM big data
TBM big data
Sliding double-windows method
Bayesian theory
Adverse geology prediction
Ensemble learning
ISSN:0886-7798
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Accurately identifying adverse geology ahead of the tunnel face in real-time is essential for guaranteeing the safe and efficient excavation of tunnel boring machines (TBMs). This study presents a novel real-time identification method for adverse geological conditions ahead of TBM tunnel face based on the stacking ensemble algorithm and Bayesian theory. Initially, a statistical analysis of 18 collapse sections from the TBM 3 section of the Yinsong Water Diversion Project in Jilin province, China, was conducted, and seven key rock-machine interaction parameters strongly correlated with adverse geological conditions were identified. Subsequently, using the sliding double-window method combined with multi-indicator judgment functions, the potential adverse geological zones were precisely identified, and an extended dataset for adverse geology identification was established. Furthermore, by integrating TBM mechanical parameters, performance parameters, geological information, and the clustering results of the Density-Based Spatial Clustering of Applications with Noise (DBSCAN), an identification model of adverse geological conditions is proposed based on the ensemble learning method. Additionally, to further improve the prediction accuracy, the Bayesian theory is introduced to refine the prediction probabilities of the proposed model. Compared with conventional machine learning classifiers, the proposed model achieves significant improvements across all evaluation metrics. The model trained on the expanded adverse geological dataset shows enhanced overall capabilities, with prediction accuracy increasing by 35.3% and Matthews correlation coefficient (MCC) improving by 19.3%. The results show that the proposed method can accurately identify the adverse geological conditions ahead and provide a conducive reference for safe excavation under complex geological conditions.
ISSN:0886-7798
DOI:10.1016/j.tust.2025.106741