危化品道路运输事故在车辆-道路-环境灾情下的演变十分复杂。为探究运输系统中不同风险因子相互作用强度和耦合特性，预防相关事故并提高运输安全，该文提出了一种基于混合模型的危化品道路运输风险演化评估方法。首先，基于事故数据库对风险因素进行辨识，确定风险耦合类型；其次，利用网络规模-相互作用度(N-K)模型，计算不同风险耦合类型的耦合度值，量化风险因素间的相互作用强度；然后，根据不同耦合类型和耦合度值构建包含耦合节点的动态Bayes网络(dynamic Bayesian network，DBN)，基于多个数据库(危化品道路运输事故综合数据库、重点路段环境监测数据库)和现有知识确定节点参数，通过考虑时间步长开展DBN分析；最后，运用风险等级指标揭示不同风险耦合类型的时间特征和风险演化潜在机理，提出风险防控策略。结果表明：多因素耦合节点的耦合程度明显高于双因素耦合节点。在双因素耦合中，驾驶员-环境耦合节点的耦合程度最高。风险耦合节点的发生概率随时间推移而逐渐增加，“违规操作”和“恶劣天气”对相应的风险耦合节点有显著影响。驾驶员因素与环境因素属于“非常强”的耦合类型。该混合模型能较准确地描述危化品道路运输事故风险演化和风险耦合的动态特征，为决策者和安全管理人员有效预防相关事故提供参考。

Objective: The evolution of accidents during road transportation of hazardous chemicals based on the conditions of vehicles, roads, and the environment is highly complex and poses a serious threat to public safety. Clarifying the intensity of the interaction and the coupling characteristics of different risk factors in the transportation system is crucial for preventing related accidents and enhancing transportation safety. However, traditional models rarely explain the mechanism of interaction between risk factors at the micro level, and the risk calculation process generally requires substantial high-quality data and information support while exhibiting limitations in handling event uncertainty. Methods: To address these issues and accurately describe the dynamic characteristics of accident risk evolution, this paper proposes a risk evolution assessment method for the road transportation of hazardous chemicals based on a hybrid model. First, risk factors are identified based on the accident database, and risk coupling types are determined. Second, the network scale-interaction degree (N-K) model is used to calculate the coupling degree for different risk types, quantifying the interaction intensity among risk factors. Thereafter, a dynamic Bayesian network (DBN) containing coupling nodes is constructed based on the different coupling types and degrees. The node parameters are determined based on multiple databases, including a comprehensive database of hazardous chemical road transportation accidents, an environmental monitoring database for key road sections, and existing knowledge. DBN analysis is performed considering time steps. Finally, the time characteristics and potential mechanisms of risk evolution across different risk coupling types are revealed using risk-level indicators, based on which risk-prevention strategies are proposed. Results: Analysis of coupling degree shows that the coupling degree of multi-factor coupling nodes is significantly higher than that of dual-factor coupling nodes. In the dual-factor coupling nodes, the degree of the driver-environment coupling node is the highest. Driver factors are more likely to couple with other factors, thereby increasing the risk of traffic accidents. The DBN analysis shows that the probability of risk coupling nodes gradually increases over time. Illegal operations and bad weather significantly affect the corresponding risk coupling nodes. Among all the risk coupling nodes, driver-environment coupling has the highest probability, indicating that driver-environment risk coupling is the most likely factor to cause accidents in the transportation of hazardous chemicals. The risk level calculation shows that the driver-environment coupling factor has the greatest impact on accident risk. Although the hazardous chemical factor does not have the highest degree of coupling, it must still be considered because of its relatively high probability of occurrence. Furthermore, although the coupling degree of multi-factor coupling nodes is higher, due to the relatively low probability of multiple factors occurring simultaneously, the risk intensity level is actually lower. Conclusions: The proposed hybrid model reveals the interaction intensity and coupling characteristics of different risk factors at the micro level. By considering the time step, the model effectively captures the evolution of accident probability driven by different types of risk coupling. Applying the hybrid model can address the challenge of assessing accident likelihood in scenarios involving the coupling of different risk factors, providing a reference to help decision-makers and safety managers effectively prevent related accidents.