Optimal Management of the Flow of Parts for Gas Turbines Maintenance by Reinforcement Learning and Artificial Neural Networks

For the maintenance of Gas Turbines (GTs) in Oil and Gas applications, capital parts are removed and replaced by parts of the same type taken from the warehouse. When the removed parts are found not completely broken, they are repaired at the workshop and returned to the warehouse, ready for future use. The management of this flow of parts is of great importance for the safe and profitable operation of a GT plant. In this chapter, we present a novel framework of part flow management, which is optimized by Reinforcement Learning (RL). The formal framework and RL algorithm account for the stochastic failure process of the involved parts. Due to the complexity of the optimization and the number of decision variables involved, we resort to action value approximation by Artificial Neural Networks (ANNs). A case study derived from a real application is worked out.