Abstract
A primary goal of maintenance is to minimize the consequences of component and system failures. Two subcategories of maintenance actions include: preventive maintenance (PM) at predetermined time intervals prior to failure and emergency repair (ER) upon failure, where the cost and downtime of ER is significantly greater than PM. Most maintenance models developed over the past several decades assume component failures are statistically independent. This assumption simplifies calculations, but is dangerous for safety critical systems that must be maintained because correlated failures can lower the mean time to failure and increase the probability of ER. This paper presents a simple method with an explicit correlation parameter to characterize the impact of correlated component failures on the optimal PM interval of a system with arbitrary structure. This method is applied to five maintenance policies, including: periodic group repair based on operation time, minimal repair with complete renewal, minimal repair with partial renewal, age replacement to minimize cost, and age replacement to maximize availability. Examples illustrate our approach identifies optimal maintenance strategies for these policies such as reward or cost per unit time, cost per renewal period, and stationary availability despite correlated failures.