Traditional risk assessment approaches mainly focus on the pre-failure scenarios with certain information. For complex systems, the scope of risk assessment needs to be extended to include the post-failure phase; because the emerging hazards of these systems cannot be wholly identified and are usually highly uncertain. Thus, resilience assessment needs to be investigated. Most of the existing literature quantify resilience based on a system's performance loss caused by disruptions. These studies fail to assess the probability of a system to sustain or restore to a normal operational state after disruptions occur, how this probability changes with time, and how fast the system can be restored. The dynamic and probabilistic characteristics of resilience must be considered in systemic resilience assessment, in which the engineered system, human and organizational factors, and external disruptions are considered. This paper aims to develop a dynamic Bayesian network (DBN)-based approach to the probabilistic assessment of the system resilience by incorporating temporal processes of adaption and recovery into the analysis of system functionality. The proposed method also provides a new way to define resilience in terms of the probability of system functionality change during and after a disruption. A case study on the Chevron refinery accident is used to demonstrate the applicability of the proposed methodology.
|Journal||Journal of Loss Prevention in the Process Industries|
|Publication status||Published - May 2020|
- Dynamic bayesian network
- Resilience assessment