A hybrid approach for optimizing deep excavation safety measures based on Bayesian network and design structure matrix

Considering the dynamic risk factors and risk situation throughout the entire deep excavation operations, timely adjustment and optimization of safety measures can enhance the practicality of construction technical plans on sites. A digital and quantitative model representing the practical risk situation of the deep excavation is urgently required for realizing the prediction, optimization, and control of the actual construction state. Thus, this research aims to propose a real-world-oriented model integrating Bayesian network (BN) and design structure matrix (DSM) for decision-making in safety risk management. First, risk factors were identified, and the BN model was established to evaluate the anti-risk ability of the construction site. Then, a multi-objective safety measure optimization model under specific constraints was established. Particularly, the DSM was adopted to express the control relationship between risk factors and safety measures. Moreover, with genetic algorithms applied, the optimal safety measure set for on-site safety risk management can be obtained. For model validation, a deep excavation project of metro construction in Wuhan, China, was selected as a case study. The hybrid optimization model showed the characters of initiative and timeliness in construction risk management. By providing the timely and optimized combination of safety measures, the dynamic decision-making approach can proactively and effectively improve the risk resistance ability of construction sites.