Real-Time Model Updating for Prediction and Assessment of Under-Construction Shield Tunnel Induced Ground Settlement in Complex Strata

Abstract

Accurate prediction of maximum ground settlement (MGS) is critical for preventing engineering accidents in tunnel construction. This study introduces a dynamic analysis approach utilizing data updating to predict MGS and evaluate the associated risks in tunneling operations under complex geological conditions. The methodology encompasses three primary components: MGS prediction; reliability assessment; and global sensitivity analysis (GSA). A refined expanded machine learning model is developed for dynamic MGS prediction, capable of effectively managing real-time data updates and identifying anomalies. Based on the dynamic prediction model, a Monte Carlo method combined with a novel functional function is used to achieve a probabilistic reliability assessment of tunnel risk. GSA using the Sobol method quantifies the impact of excavation parameters on MGS. The results show that the proposed approaches have the potential for MGS prediction and tunnel risk assessment in complex strata. This study advances dynamic MGS probabilistic analysis approach in complex strata.