Flight schedule design and fleet assignment are the two main elements of the airline scheduling process, which have the highest effect on cost and revenue. Although mixed-integer linear programming models were developed for integrated schedule design and fleet assignment, it has been shown that this approach was not efficient for large-scale models. Therefore, this paper aimed at developing a parallel master–slave Genetic Algorithm (PMS-GA) for solving the integrated flight schedule design and fleet assignment problem with demand recapture, particularly for large-scale problems. The integrated schedule design and fleet assignment problem was solved by the master GA, while the slave GA nested inside the master GA solved passenger flow adjustment problem. Considering the complexities of a large-scale integrated problem, we (1) proposed an innovative approach for creating feasible suboptimal initial population, (2) developed customized genetic operators to improve the performance of the PMS-GA compared to the conventional GAs, and (3) implemented migration and repopulation to prevent premature convergence. PMS-GA was tested on seven models with small-, medium-, and large-scales, and the results were compared with the gold-standard mixed-integer linear programming in terms of cost and runtime. The comparative study showed that the PMS-GA achieved suboptimal solutions with costs only 1.8% to 3.0% different than the optimal solution for medium- and large-scale models. However, these solutions were obtained in significantly shorter runtimes (over 500% to 1000%) compared to the mixed-integer linear programming. Also, the results showed that in contrast to the mixed-integer linear programming approach, runtimes of the proposed PMS-GA are highly predictable as a function of the problem size. Our results showed the importance of PMS-GA for integrated schedule design and fleet assignment, particularly for solving large-scale re-scheduling problems in a short time.

航班时刻表设计和机队分配是航空公司时刻表编制过程中的两个主要要素，对成本和收入的影响最大。尽管混合整数线性规划模型是为集成计划设计和车队分配而开发的，但事实表明，这种方法对于大规模模型而言效率不高。因此，本文旨在开发一种并行的主从遗传算法（PMS-GA），以解决具有需求重新捕获的综合飞行计划设计和机队分配问题，特别是对于大规模问题。集成的时间表设计和车队分配问题由主GA解决，而从GA嵌套在主GA中解决了客流调整问题。考虑到大规模综合问题的复杂性，我们（1）提出了一种创新的方法来创建可行的次优初始种群，（2）开发了定制的遗传算子以与传统GA相比提高PMS-GA的性能，并且（3）实施迁移和种群重组以防止过早收敛。PMS-GA在七个模型上进行了小，中，大规模的测试，并且在成本和运行时间方面，将结果与金标准的混合整数线性编程进行了比较。对比研究表明，PMS-GA实现了次优解决方案，其成本与中型和大型模型的最优解决方案仅相差1.8％至3.0％。但是，与混合整数线性编程相比，这些解决方案是在明显更短的运行时间（超过500％至1000％）中获得的。也，结果表明，与混合整数线性规划方法相比，所提出的PMS-GA的运行时间可根据问题大小进行高度预测。我们的结果表明，PMS-GA对于集成计划设计和机队分配非常重要，特别是对于在短时间内解决大规模重新计划问题。