考虑多约束的MOJ调度问题

统筹考虑晶圆加工过程中的多品种、p-s-d(past-sequence-dependent)换模时间及衰退效应等约束特征,以总加权提前/拖期惩罚成本最小为优化目标,建立了单机MOJ(multiple orders per job)调度数学规划模型.在此基础上,对决策变量进行分离,提出具有双层嵌套编码机制的改进型遗传蚁群调度算法.该算法将遗传算法融合到动态自适应蚁群算法的每一次迭代过程中,并为有效提高算法的收敛性能,引入ATCS(apparent tardiness cost with setups)修正准则.最后,仿真实验结果表明,该算法是有效、可行的.

Scheduling Multiple Orders per Job with Various Constraints

Taking a comprehensive consideration of the characteristics of multiple product types， the past-sequence-dependent (p-s-d) setup time and the deterioration effects constraints in processes of wafer fabrication， with an objective function of minimizing total weighted earliness-tardiness penalties cost， a mathematical programming model of scheduling multiple orders per job (MOJ) in a single machine was built. On this basis， the decision-making variables were separated， and a modified genetic algorithm-ant colony optimization (MGA-ACO) algorithm adopting two-level encoding mechanism was put forward. Genetic algorithm was converged to the process of dynamic and adaptive ant colony iterations. To improve the algorithm convergence performance， a modified rule of apparent tardiness cost with setups (ATCS) was applied. Finally， the simulation results indicated that the developed algorithm is valid and feasible.