Using Bayesian network for safety risk analysis of diaphragm wall deflection based on field data

Highlights: • A BN model for safety risk analysis of diaphragm wall deflection is developed. • Current field data is used to compute the dynamic dependencies of risk variables. • Historical field data is used to determine the initial probabilities of root nodes. • The prediction results of BN model are verified in a real metro construction project. This study proposes a risk analysis model based on the Bayesian network for diaphragm wall deflection to ensure safety during excavation. Compared with the traditional methods, such as fault tree analysis, the proposed approach can more accurately illustrate the dynamic and updated features of hydrogeological design and construction variables as construction progresses on the basis of metro construction field data, which can address the risks in uncertain environments and achieve dynamic control during the entire construction process. The procedures include risk identification, sample collection, model building and validation, risk analysis and reasoning, and real-time control measures. The proposed model was applied in the metro station construction project, in Wuhan Metro Line 7. Results showed that the proposed risk analysis method can achieve real-time risk analysis, which includes pre-accident, during-construction, and post-accident control, to ensure construction safety at the station.