Reinforcement Learning for Mission Reliability Based Selective Maintenance Optimization

The majority of Selective Maintenance Optimization (SMO) techniques necessitate a longer computation time for the determination of the maintenance strategy, a characteristic that is considered undesirable. The advent of machine learning presents opportunities for the exploration of innovative methodologies that have the potential to substantially reduce computational time in SMO. This paper proposes a novel Reinforcement Learning (RL) based methodology for SMO. A temporal difference learning algorithm—Q-Learning is used to solve this problem where the agent chooses a policy at the end of an epoch based on the updated Q-Values. Different heuristics are embedded with the methodology to effectively determine the optimal policy that results in achieving desired mission reliability, of which one smartly reduces the solution space, and the other aids in increasing the agent’s intelligence based on the reward policy. The objective of the maintenance optimization problem is to achieve the desired system mission reliability while consuming minimum resources. The reward function is defined such that the agent will learn and then determine the best strategy considering all the constraints. The efficiency of the developed algorithm is demonstrated by applying it to a benchmark coal transportation system. Results accentuate the supremacy of the developed RL-based algorithm over the commonly used methods like the enumeration approach and genetic algorithm based approach.