Optimal resilience-based restoration of a system subject to recurrent dependent hazards

Many engineered systems are faced with multiple natural and manmade hazards during operation. Since such hazards may cause system disruptions (e.g., catastrophic failure and performance deterioration), timely and effective restoration needs to be conducted to minimize the negative impacts of the hazards. In practice, these hazards are recurrent and often interact dynamically with each other. This causes a major challenge in assessing the performance of the systems and determining the optimal restoration policy from a holistic perspective. In this paper, we model the availability and resilience of a system subject to recurrent and dependent hazards and develop the optimal restoration policies for decision makers with different risk preferences. Particularly, we first model the random occurrence times and severities of a set of hazards whose occurrence frequencies and severities are dynamically dependent. Next, we derive the analytical expressions for the system's availability metrics. Third, we investigate the optimal restoration policies for both risk-neutral (RN) and risk-averse (RA) decision makers by solving two related optimization problems. A case study on the pipelines in the Mid-Atlantic area of the U.S. is provided to illustrate the practical values of the proposed availability assessment and resilience optimization methods.