关键词: 交通运输工程；液化石油气；风险分析；泄漏事故；动态贝叶斯网络；演变过程

Abstract: To reveal the temporal evolution patterns of risks in liquefied petroleum gas (LPG) railway tanker transportation, a dynamic Bayesian network risk assessment model integrated with temporal dimensions was proposed. By employing a bow-tie model to unify fault tree analysis and event tree analysis method, a static risk topology structure for transportation leakage accidents was constructed, incorporating temporal parameters to characterize dynamic risk propagation. Based on GeNIe platform, numerical simulations and backward reasoning were conducted. Research results show that: ①the probability of leakage accident in initial state is 13.43%, escalating to 48.13% after 4 h without intervention; ②explosion accidents are more likely to occur after leakage, and the risks exhibit nonlinear escalation, with accident probability occurring within 4 hours increasing from 4.18% to 14.99%;③safety accessory/equipment failures and abnormal tanker pressure emerge as critical causative factors of leakage. The research demonstrates that dynamic Bayesian networks can effectively quantify time-varying risks of hazardous material transportation systems, providing theoretical support for dynamic early warning and intervention decision-making in leakage accidents.

Key words: traffic and transportation engineering; liquefied petroleum gas; risk analysis; leakage accident; dynamic Bayesian network; evolutionary process