Optimal maintenance of a critical infrastructure subject to dependent recurrent hazards

Natural and human-made hazards deteriorate critical infrastructures and cause failures. To reduce losses, operators perform preventive maintenance. However, such hazards are often recurrent and interact dynamically with each other, complicating the preventive maintenance policy optimization. In this study, we propose a framework to model system availability and resilience with one or more performance measures under dependent recurrent hazards, and to derive optimal preventive maintenance for varying risk preferences. Specifically, we characterize hazards whose occurrence frequencies and severities are interdependent, model system availability and resilience under these interactions, and formulate two optimization problems that yield policies for risk-neutral and risk-averse decision makers. Efficient algorithms are developed to evaluate availability and sample representative hazard scenarios while searching for optimal policies. A case study on the underground pipeline system in Pennsylvania and New York is provided to illustrate the application of the proposed resilience assessment and maintenance optimization methods.