Applications of Reinforcement Learning for maintenance of engineering systems: A review

Abstract

Nowadays, modern engineering systems require sophisticated maintenance strategies to ensure their correct performance. Maintenance has become one of the most important tasks of the systems lifecycle. This paper presents a literature review of the application of Reinforcement Learning algorithms for the maintenance of engineering systems. Reinforcement Learning-based maintenance has been classified regarding four types of system: transportation systems, manufacturing and production systems, civil infrastructures, power and energy systems, and other systems. Based on the literature review, this paper includes an overall analysis of the current state and a discussion of main limitations, challenges, and future trends in this field. A summary table is provided to present clearly the most important references. This research work demonstrates that Reinforcement Learning algorithms have a great potential for generating maintenance policies, outperforming most conventional strategies.