Methods for Improving Robustness and Recovery in Aviation Planning.

Abstract

In this dissertation, we develop new methods for improving robustness and recovery in aviation planning. In addition to these methods, the contributions of this dissertation include an in-depth analysis of several mathematical modeling approaches and proof of their structural equivalence. Furthermore, we analyze several decomposition approaches, the difference in their complexity and the required computation time to provide insight into selecting the most appropriate formulation for a particular problem structure. To begin, we provide an overview of the airline planning process, including the major components such as schedule planning, fleet assignment and crew planning approaches. Then, in the first part of our research, we use a recursive simulation-based approach to evaluate a flight schedule's overall robustness, i.e. its ability to withstand propagation delays. We then use this analysis as the groundwork for a new approach to improve the robustness of an airline's maintenance plan. Specifically, we improve robustness by allocating maintenance rotations to those aircraft that will most likely benefit from the assignment. To assess the effectiveness of our approach, we introduce a new metric, maintenance reachability, which measures the robustness of the rotations assigned to aircraft. Subsequently, we develop a mathematical programming approach to improve the maintenance reachability of this assignment. In the latter part of this dissertation, we transition from the planning to the recovery phase. On the day-of-operations, disruptions often take place and change aircraft rotations and their respective maintenance assignments. In recovery, we focus on creating feasible plans after such disruptions have occurred. We divide our recovery approach into two phases. In the first phase, we solve the Maintenance Recovery Problem (MRP), a computationally complex, short-term, non-recurrent recovery problem. This research lays the foundation for the second phase, in which we incorporate recurrence, i.e. the property that scheduling one maintenance event has a direct implication on the deadlines for subsequent maintenance events, into the recovery process. We recognize that scheduling the next maintenance event provides implications for all subsequent events, which further increases the problem complexity. We illustrate the effectiveness of our methods under various objective functions and mathematical programming approaches.