Reinforcement learning-based flow management of gas turbine parts under stochastic failures

For 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 broken, they are repaired at the workshop and returned to the warehouse, ready to be used in future maintenance. The management of this flow is of great importance for the profitability of a GT plant. In this paper, we adopt a previously developed formalized framework of the part flow and reinforcement learning (RL) to optimize part flow management. The formal framework and RL algorithm are extended to account for the stochastic failure process of the involved parts. An application to a scaled-down case study derived from an industrial application is illustrated.