To enhance the safety risk management and control capabilities for TBM (Tunnel Boring Machine) construction in hydraulic tunnels, this study conducts a correlation analysis and dynamic evolution study of safety risks. Data were collected through multiple channels, including a literature review, on-site records, and expert interviews. Grounded theory was employed for three-level coding to initially identify risk factors, and gray relational analysis was used for indicator optimization, ultimately establishing a safety risk system comprising 5 categories and 21 indicators. A multi-level hierarchical structure of risk correlation was established using fuzzy DEMATEL and ISM, which was then mapped into a Bayesian network (BN). The degree of correlation was quantified based on probabilistic information, leading to the construction of a risk correlation analysis model based on fuzzy DEMATEL–ISM–BN. Furthermore, considering the risk correlations, a safety risk evolution model for TBM construction in hydraulic tunnels was developed based on system dynamics. The validity of the model was verified using the AY project as a case study. The results indicate that the safety risk correlation structure for TBM construction in hydraulic tunnels consists of 7 levels, with the closest correlation found between “inadequate management systems” and “failure to implement safety training and technical disclosure”. As the number of interacting risk factors increases, the trend of risk level evolution also rises, with the interrelations within the management subsystem being the key targets for prevention and control. The most sensitive factors within each subsystem were further identified as adverse geological conditions, improper construction parameter settings, inappropriate equipment selection and configuration, weak safety awareness, and inadequate management systems. The control measures proposed based on these findings can provide a basis for project risk prevention and control. The main limitations of this study are that some probability parameters rely on expert experience, which could be optimized in the future by incorporating more actual monitoring data. Additionally, the applicability of the established model under extreme geological conditions requires further verification.
Nie et al. (Wed,) studied this question.