Bayesian estimation and consequence modelling of deliberately induced domino effects in process facilities

Abstract Process facilities handling hazardous chemicals in large quantities and elevated operating conditions of temperature/pressure are attractive targets to external attacks. The possibility of an external attack on a critical installation, performed with an intention of triggering escalation of primary incidents into secondary and tertiary incidents, thereby increasing the severity of consequences needs to be effectively analysed. A prominent Petrochemical Industry located in Kerala, India was identified for studying the possibility of a deliberately induced domino effect. In this study, a dedicated Bayesian network is developed to model the domino propagation sequence in the chemical storage area of the industry, and to estimate the domino probabilities at different levels. This method has the advantage of accurately quantifying domino occurrence probabilities and identifying possible higher levels of escalations. Moreover, the combined effect from multiple units can be modelled easily and new information can be added into the model as evidences to update the probabilities. Phast (Process hazard analysis) software is used for consequence modelling to determine the impact zones of the identified primary and secondary incidents. The results of the case study show that such analyses can greatly benefit green field and brown field projects in determining the appropriate safety and security measures to be implemented or strengthened so as to reduce its attractiveness to external threat agents.