Gestion des avions et des equipages durant le jour d'operation.

摘要

Airlines are faced with the problem of producing aircraft flight itineraries and crew rotations to provide scheduled service, while maximizing profits. This type of problem is referred as Planning problem and it is solved in advance, i.e., few days, weeks or months before the day of operation. Another type of problem arrives when perturbations occur on the day of operation. In such a case the flight schedule may become infeasible and must be updated. The problem must be solved in real-time while the airline operations are in progress.;For this type of problem, which we refer to as the Day of Operation Scheduling (DAYOPS) problem, we suggest three approaches that solve aircraft and crew assignment simultaneously. The first approach proposes a formulation designed to determine a new flight schedule that fits with the existing aircraft assignment, maintenance schedule, crew schedule and passenger connections. The dual model can be formulated as a network flow model. Using this approach, it is possible to solve in real time a special case of DAYOPS problems at the largest airlines. The proposed model considers relatively small irregularities that permit no changes in aircraft itineraries and crew rotations. In addition, it may be possible to embed the proposed model in more sophisticated operational or planning systems, such as a two-level optimisation model for aircraft and crew schedules.;The second approach is more general and permits changing aircraft itineraries, crew rotations and the planned schedule. Each part of the optimisation process, as aircraft, pilots and flight attendants, is separately solved. We propose mathematical formulations that relate models in this sequential approach. Finally, in the third approach, the Benders decomposition is used to separate integral multi-commodity flow formulation in two parts where the aircraft assignment problem is the master problem and the crew assignment problem is the sub-problem. Each of these parts is solved separately by Dantzig-Wolfe decomposition where we define one network for every commodity. In each pail we use a Branch and Bound technique to find an integer solution. The last two approaches have bigger responding time during an airline operation but they are able to solve bigger irregularities problem. Usually, the bigger irregularities do not need short responding time because they could be known longer in advance.;The principal contributions of this dissertation are: identifying and defining three approaches for solving a DAYOPS problem; solving efficiently a special case of a DAYOPS problem; finding the linking relations between the principal actors in an operational problem; formulating an integrated optimisation model; developing a sophisticated solution method to solve those integrated model. We believe that the proposed ideas are promising and that further research could produce an operational schedule that considerably reduces the cost to the airline, a potentially decisive factor in an extremely competitive airline market.