关键词: 不正常航班恢复, 粒子群算法, 联合恢复, 机组竞拍

Abstract:

Irregular flights deviate from their schedules， which causes a great of financial losses for airlines. In the process of recovery， aircraft and crew resources play a vital role and can be controlled by the airlines. However， due to the complex and stringent constraints in the aviation industry， previous studies have primarily focused on single-resource independent recovery or simple integrated cases， overlooking the importance of crew satisfaction in service quality and airline reputation. To address this gap and better reflect real-world scenarios， an integrated aircraft and crew recovery model that incorporates a crew satisfaction evaluation mechanism is proposed in this paper. In the modeling， crews are given the opportunity to bid for the earliest off-duty task in the event of temporary airport closure. To achieve the best possible recovery scheme， five recovery options are simultaneously considered： flight delay， flight cancellation， crew exchange， standby crew utilization， and cruise speed increase. Although this leads to increased problem dimension and solution difficulty， it ensures a comprehensive approach to recovery. Given the high constraints and coupled solution difficulties of the model， an ad-hoc intelligent algorithm based on Particle Swarm Optimization （PSO） is designed. A novel tripartite competition strategy and model information-guided method are innovatively incorporated to enhance the performance of the algorithm. Specifically， tripartite competition strategy involves randomly selecting a particle from three subpopulations to generate “win”， “lo{s}_{\mathrm{1}}” and “lo{s}_{\mathrm{2}}” based on their fitness and feasibility values. These particles are then updated using different evolutionary strategies. The information-guided method facilitates neighbor searching for feasible and optimal solutions， thereby improving solution performance and avoiding algorithmic stagnation. Additionally， a set of infeasible solution repair mechanisms， such as boundary and extension handlings， are designed to increase the probability of finding feasible solutions.Finally， to validate the effectiveness of our proposed solving algorithm， it is compared with two classical algorithms （PSO and ABC）， two novel algorithms （MSEFA and MSRCS）， and three algorithms used for aircraft and crew recovery problems （GA_arp， SA_arp， and IFWA_arp） on three different-scale instances from Shenzhen Airlines and the 2017 Competition. The results， evaluated from both computational and statistical perspectives， demonstrate that our proposed algorithm outperforms the competitors in terms of precision in solving the recovery problem， irrespective of the problem scale. Furthermore， it is recommended to increase the iteration number to more than 3000 for large-scale instances， which can yield approximately 10% cost savings for recovery operations. To sum， an effective intelligent tool for addressing complex aircraft and crew integrated recovery problems is offered in this study， reducing labor-intensive and experience-dependent challenges associated with manual recovery processes.

Key words: irregular flight recovery, particle swarm optimization, integrated recovery, crew bidding