Dynamic risk assessment of chemical process systems using Bayesian network

Background and aims: Process systems due to processed under severe operational conditions and deal with large amounts of flammable and explosive materials have always led to many catastrophic accidents. Risk assessment is a useful tool for designing effective strategies for preventing and controlling these accidents. Conventional risk assessment methods have major deficiencies, including uncertainty in the obtained results and the completely static nature, therefore, the present study is aimed at applying a dynamic and quantitative approach to assess the safety risks of city natural gas pressure regulating stations. Methods: First, according to the standard of the Total Company (GS EP SAF 253), the reference (credible) event scenario was determined, then a qualitative, quantitative and dynamic modeling of the cause-consequence accident scenario model using Bayesian Network (BN) is provided and next, deductive and abductive probabilistic reasoning are conducted by means of constructed BN model. PHAST 7.11 program is employed to modeled and evaluated of different consequences of the scenario. Finally, the risk of accident scenario consequences was calculated, evaluated and updated. Results: 43 root events in occurrence of the credible event scenario of the gas stations were identified. Among the identified causes, the human failures (85%), process failures (10%) and mechanical failures (5%) had the highest contributing to occurrence in the accident scenario, respectively. Occurrence probability of the scenario is determined 7.11 ×10-2. Safety barriers, especially emergency shutdown valves (ESD), had a significant role in reducing the consequences severity. The risk of all three of consequences including jet fire, flash fire and vapor cloud explosion is located in unacceptable area. Conclusion: The use of BNs provides a comprehensive qualitative, quantitative and dynamic graphical modeling of the accident scenario. The abductive reasoning ability of these networks is capable to reducing the uncertainty and updating the probability of occurrence of root events and final the consequences. Using BNs along with consequences modeling, leads to a slightly more dynamic, precise and practical risk assessment in process plants.