Quantitative risk assessment of EO leakage integrating KG, BN, and CFD simulation

Abstract Ethylene oxide (EO), a crucial derivative of ethylene, involves high temperature and pressure as well as flammable, explosive, and toxic materials. This complexity predisposes its production to severe industrial accidents. To systematically identify the risk factors that may lead to EO accidents, this study conducted an integrated analysis by leveraging a knowledge graph (KG), a Bayesian network (BN), and computational fluid dynamics (CFD) simulation. First, the key risk factor nodes leading to leakage were identified by using KG. Subsequently, a BN was constructed, which enabled the clarification of typical EO leakage scenarios. Then, for the identified typical leakage scenarios, a CFD geometric model was developed. This model was employed to simulate the dispersion patterns following an EO leak under different conditions. Finally, by integrating the results from the BN and CFD analyses, the working conditions with the highest risk level were identified. The results indicated that leakage caused by chemical corrosion or physical erosion of reactors or pipelines presented the highest quantified risk value. This scenario exerts a significant impact on the safe production of EO, necessitating prioritized and focused inspection. Through the integrated KG–BN–CFD methodology, this study has effectively elucidated the disaster-causing mechanisms and evolution patterns of EO leakage accidents. It thereby provides a theoretical foundation and practical guidance for the precise identification of risks, the formulation of targeted prevention and control measures, and the enhancement of inherent safety levels in EO production facilities.