Development of a Bayesian Network–Based Integrated Model for Opening‐Related Fall Risk Assessment in Construction

Opening‐related falls remain a critical safety concern in construction, particularly around floor openings and elevator shafts where risk is jointly influenced by opening configurations, protective conditions, worker proximity, and unsafe behaviors. However, existing assessment approaches often evaluate these factors statically and have limited capability to connect design stage opening hazards with evolving worker conditions during construction. This study proposes an integrated probabilistic framework combining building information modeling (BIM), Bayesian networks (BNs), dynamic Bayesian networks (DBNs), and computer vision (CV) for dynamic assessment of opening‐related fall risk across the design and construction phases. At the design stage, opening geometry, type, edge protection, and worker‐to‐opening distance derived from 2D BIM are integrated with expert elicited probabilities in a static BN to quantify inherent fall risk. This design stage risk is subsequently transferred to a DBN as baseline information, while CV‐derived worker observations are incorporated as temporal evidence to update fall risk probabilities during construction. Three complementary case studies are presented: (1) static BN assessment of inherent risk associated with building openings; (2) controlled evaluation of temporal risk evolution near an opening; and (3) field‐based assessment using construction site observations. The field application further compares DBN predictions with static BN and expert rule‐based assessments. CV‐derived width‐to‐height ratios showed close agreement with manual measurements (MAE = 0.0152, RMSE = 0.0167, MAPE = 3.744%, and R ²  = 0.9927). The results demonstrate that the proposed framework links persistent opening‐related hazards with changing worker states and captures the temporal evolution of fall risk through sequential probabilistic updating. This framework provides a risk‐based foundation for proactive monitoring and decision making to enhance safety during work near hazardous openings in construction.