Customer-oriented group maintenance model for unreliable service system

This study considers a service system with multiple independent servers operating in parallel and a single Markovian queue. The servers are unreliable with identically exponentially distributed failure times but no server failures allowed during maintenance. The repair time is also assumed to follow exponential distribution. Accordingly, we formulate this model as a continuous time Markov decision process. Group maintenance policies are developed based on the number of customers dynamically in the queue, and are mathematically proved to have the property of threshold structure.