Delay Averse Crew Pairing Optimization

Abstract

Airline operations involve many complex and interdependent decisions. These include flight scheduling, fleet assignment, maintenance routing, and crew scheduling. Crew scheduling is particularly important, as crew costs are the second-largest expense in airline operations. In practice, delays are common and can cause major disruptions to crew schedules. Most crew pairing models are deterministic and aim to minimize planned costs. However, they often ignore operational uncertainties, such as delays. As a result, they may perform poorly in real operations. This is critical because crew payments are based on actual flight times. This thesis addresses the delay averse crew pairing problem. We propose an optimization model with a delay cost component to create more reliable pairings. The model is based on a duty network structure, and utilizes the pay-and-credit scheme to account for crew payments. We solve the model using Lagrangian relaxation and perform extensive testing on real data. We compare the delay averse model to the nominal case and conduct sensitivity analysis to demonstrate how changes in cost parameters and delay levels affect the solutions. Finally, we evaluate the solutions using simulation based on historical flight delay data. The results show that the delay averse model creates more reliable pairings. These pairings are better at absorbing delays and reducing total propagated delays.