Resilience assessment of critical infrastructures using dynamic Bayesian networks and evidence propagation

The proper functioning of critical infrastructures is vital for supporting the economy and social welfare worldwide. Therefore, several methods were developed to assess the resilience of such systems in the face of disruptive events. This work proposes a novel probabilistic approach to the resilience assessment of critical infrastructures using a dynamic Bayesian network (DBN) to model resilience curves and cumulative impact in the face of failures. The DBN's structure is based on the physical connections of the system, allowing for a more generalist methodology. Additionally, evidence propagation allows for a scenario-driven approach. Any failure and repair scenario is modelled as evidenced in the DBN, allowing all other nodes' conditional probabilities to be updated accordingly. An Electric Power Distribution System is used to validate the methodology, and results show the ability of the DBN coupled with evidence propagation to assess the resilience of complex systems. Different failure scenarios and restorative actions are considered, resulting in comparative results which can guide decisions and investments in the system.