基于离散教与学算法的分布式预制流水车间调度研究

A Discrete Teaching and Learning Algorithm for Flow Shop Scheduling of Distributed Prefabricated Components

摘要:

分布式工厂生产形式对提高预制构件生产效率、保证订单按时交付、降低企业拖期交货惩罚费用具有重要的意义。因此针对分布式预制构件流水车间调度问题，以最小化订单总拖期惩罚为目标建立了数学优化模型，并基于双层整数编码方式提出了一种离散教与学算法（DTLBO）。在算法初始化阶段，采用启发式规则和随机生成融合策略改善初始解的质量，进而增加算法的寻优效率；在教学阶段，结合问题模型特点，设计了顶层替换、底层替换两种邻域构造，促进教师解对学生解的引导优化；在学习阶段，通过变异算子和交叉算子让学生解之间相互学习更新，进一步提升算法的局部开发和全局探索能力。试验结果表明，与遗传算法和变邻域搜索算法对比，提出的DTLBO算法具有更好的求解性能和鲁棒性。最后与实际生产过程常用的经验启发式调度方法相比，提出算法在目标值上表现出不低于10%的平均改进率，有望显著增加预制构件制造企业净利润并提高客户满意度，能够为企业管理者提供更佳合理的生产调度方案。

Abstract:

Distributed factory production is of great significance to improve the production efficiency of prefabricated components, ensure the on-time delivery of orders, and reduce the penalty cost of delayed delivery. Therefore, in order to minimize the penalty of total order delay, a mathematical optimization model is established for the scheduling problem of distributed prefabricated component flow shop, and a discrete teaching-learning based optimization (DTLBO) is proposed based on double-layer integer coding. In the initial stage of the algorithm, heuristic rules and random generation fusion strategy are used to improve the quality of the initial solution, so as to increase the optimization efficiency of the algorithm; in the teaching stage, combined with the characteristics of the problem model, two kinds of neighborhood structures: top-level replacement and bottom-level replacement, are designed to promote the guidance and optimization of the teacher's solution to the student's solution; in the learning stage, the mutation operator and crossover operator are used to let students learn and update each other, so as to further improve the local development and global exploration ability of the algorithm. Experimental results show that compared with genetic algorithm and variable neighborhood search algorithm, the proposed DTLBO achieves better solution quality and robustness. Finally, compared with the empirical heuristic scheduling method commonly used in the actual production process, the proposed algorithm shows an average improvement rate of no less than 10% on the target value, which is expected to significantly increase the net profit of the prefabricated component manufacturing enterprise and improve customer satisfaction, and can provide a better and reasonable production scheduling scheme for enterprise managers.