Probabilistic multi-scale modeling of interdependencies between critical infrastructure systems for resilience

Abstract

The prevalence of aging infrastructure and an increase in cascading failures have highlighted the need to focus on building strong, interdependent infrastructure systems to increase resilience. To understand the ways infrastructure systems depend on one another, we define three comprehensive interdependency types – service provision, geographic, and access for repair. We propose a methodology to model interdependencies probabilistically using a novel Bayesian network approach. By understanding how these interdependencies affect the fragility of overall systems, infrastructure owners can work towards creating more resilient infrastructure systems that sustain less damage from natural hazards and targeted attacks, and restore services to communities rapidly. Generalized expressions to create the multi-scale Bayesian network model accounting for each interdependency type are presented and applied to a real interdependent water, power, and gas network to demonstrate their use. These models enable us to probabilistically infer which interdependencies have the most critical effects and prioritize components for repair or reinforcement to increase resilience.