Coupling Risk Analysis of Roadway Tunnel Fires Using Dynamic Bayesian Network and N-K Model

To investigate the multifactor coupling mechanisms in roadway tunnel fire accidents, this study develops a quantitative risk analysis framework integrating the N-K model and a dynamic Bayesian network (DBN). Historical accident statistics are analyzed to identify coupling patterns among human, tunnel, vehicle, and management subsystems. The N-K model quantifies coupling strength and interaction mechanisms, while the integrated N-K–DBN model enables dynamic probabilistic assessment, diagnostic evaluation, and sensitivity analysis. Using empirical data reduces reliance on expert-based assessments and may improve the transparency and consistency of the evaluation within the adopted coding and modeling assumptions. The findings indicate that human–vehicle and human–tunnel–vehicle couplings exhibit the highest probabilities and are, within the present modeling framework, most strongly associated with traffic accidents, vehicle categories, and emergency management proficiency. Sensitivity analysis provides model-based insights for formulating operation and maintenance strategies under the adopted assumptions. During early operation, the results suggest that strengthening traffic control and improving emergency response capability may help reduce accident likelihood. In later stages, maintaining the stable operation of tunnel systems may help limit coupling risks. Overall, this study provides a model-based quantitative reference for understanding dynamic coupling paths and for supporting fire-safety-related decision making in roadway tunnel systems.