Heuristic and exact techniques for aircraft maintenance scheduling

Aircraft's availability is certainly one of the most important features of modern avionic industry. High availability can only be obtained with very efficient maintenance cycles. These cycles, in turn, are extremely expensive in terms of tools and personnel. This article describes the main features of an aircraft maintenance cycle contrasting it with other similar scheduling tasks. After that, it demonstrates how to model a maintenance cycle to enable a symbolic (mathematical) analysis, and it shows how to create a plan with both heuristic and symbolic (exact) techniques. Heuristic techniques show high efficiency and scalability, but they deliver sub-optimal results, and are then unable to minimize maintenance costs. Exact techniques are able to find optimal solutions even for very constrained tasks. Unfortunately, even the most efficient exact strategy is not able to deal with real (complete) maintenance problems. As a consequence, heuristic and exact strategies are used together to trade-off the accuracy of the result with the capacity of the considered problems. From the experimental point of view, the article reports data on real maintenance tasks coming from the avionic industry. It describes how to discover and to correct error coded in the original database, which all previous manual analyses were unable to reveal. The final scheduling shows consistent improvements against the original manual planning adopted on the field so far.