Airline maintenance scheduling (AMS) studies how plans or schedules are constructed to ensure that a fleet is efficiently maintained and that airline operational demands are met. Additionally, such schedules must take into consideration the different regulations airlines are subject to, while minimising maintenance costs. In this thesis, we study different formulations, solution methods, and modelling considerations, for the AMS and related problems to propose two main contributions.

First, we present a new type of multi-objective mixed integer linear programming formulation which challenges traditional time discretisation. Employing the concept of time intervals, we efficiently model the airline maintenance scheduling problem with tail assignment considerations. With a focus on workshop resource allocation and individual aircraft flight operations, and the use of a custom iterative algorithm, we solve large and long-term real-world instances (16000 flights, 529 aircraft, 8 maintenance workshops) in reasonable computational time. Moreover, we provide evidence to suggest, that our framework provides near-optimal solutions, and that inter-airline
cooperation is beneficial for workshops.

Second, we propose a new hybrid solution procedure to solve the aircraft recovery problem. Here, we study how to re-schedule flights and re-assign aircraft to these, to resume airline operations after an unforeseen disruption. We do so while taking operational restrictions into account. Specifically, restrictions on aircraft, maintenance, crew duty, and passenger delay are accounted for. The flexibility of the approach allows for further operational restrictions to be easily introduced. The hybrid solution procedure involves the combination of column generation with learning-based hyperheuristics. The latter, adaptively selects exact or metaheuristic algorithms to generate columns. The five different algorithms implemented, two of which we developed, were collected and released as a Python package (Torres Sanchez, 2020). Findings suggest that the framework produces fast and insightful recovery solutions.