|dc.description.abstract||Flight schedules are often affected by unexpected interruption events, causing aircraft to be unable to take off or land on time, also resulting in connection delays as well as later flights. Taiwan is located in the Western Pacific on the path of typhoons, so one of the worst and the most frequent interruption events are typhoons. In practice, when it is known that a typhoon is approaching, decision makers make adjustments to flight schedules and fleet routes based on experience and the aviation weather forecast data, which is neither efficient nor effective. Moreover, the manual adjustment of airline network operations may not be considered systematically, usually leading to feasible but inferior reassignments. The primary focus in past studies of flight scheduling problems has been on the planning stage. However the planning problem, which does not need to be solved immediately, is different from real-time flight schedule adjustment problems in response to typhoon events. Although there have been some studies discussing real-time flight schedule adjustment problems, the problem of passenger transportation has been neglected and the duration of the determined disruption has been over-simplified. In fact, the duration of typhoon is uncertain, so the current models cannot be applied. Therefore, in this study, we develop a deterministic flight rescheduling model and a stochastic flight rescheduling model, aimed at the minimization of the total operating cost. These models should be useful tools to assist decision makers to make adjustment to flight schedules and fleet routes in response to typhoon disruption events.
This dissertation is divided into two parts. In the first essay, we consider the practices, current operations and limitations of an airline carrier for the construction of a deterministic flight rescheduling model dependent upon a precise typhoon disruption period. In the second essay, we construct a stochastic flight rescheduling model given an uncertain typhoon disruption period. In addition, a dynamic application using a multi-stage decision-making process is adopted to illustrate the flexibility and practicality of the two models. Network flow techniques and mathematical programming are utilized to develop all the models, coupled with the constraints used in practice. All the models are formulated as integer network flow problems with side constraints, and are characterized as NP-hard. To efficiently solve the realistically large problems that occur in practice, a solution algorithm based on the divide and conquer approach is developed. The models are written using the C++ computer language, coupled with the CPLEX mathematical programming solver, to solve the problems. Finally, numerical tests are performed, and some conclusions and suggestions for future research are given.||en_US|