Resilience assessment of process industry facilities using dynamic Bayesian networks

Facilities in the process industry manufacture, store, and transfer hazardous materials which are explosive, flammable, and toxic. In industrial facilities, units are interdependent and closely located, which makes them vulnerable to cascading accidents. An undesired disruption to one unit might propagate to others, leading to more performance losses and more efforts to repair the damaged facility. For interdependent facilities which might require a lot of effort to rebuild, resilience assessments through quantifiable performance metrics can be used to consider adaptation and restoration during the post-disruption stage and thereby account for the impacts of disruptions on the performance of facilities during both pre-disruption and post-disruption stages. This study proposes a framework to measure the resilience of facilities vulnerable to cascading accidents in the process industry. Dynamic Bayesian network is applied to model the possible spatial and temporal evolution scenarios of cascading accidents. The uncertainties in evolution paths during the escalation of cascading accidents are considered. A case study of a storage tank farm is applied to illustrate the application of the framework. The effects of protection measures on the resilience to fire hazard are evaluated through cost-benefit analysis to determine the optimal protection strategy of the tank farm. A sensitivity analysis based on the case study is conducted to study the influence of critical parameters on resilience assessment.