为有效解决晶圆加工过程中带换模时间、品种间晶舟分配的不确定性以及参数调整等多重加工前约束的单机单作业多订单MOPJ(multi-order-per-job)调度问题,对问题域进行描述,以订单总完成时间最小为优化目标,建立数学规划模型.给出求解较优调度解的定理,并提出具有双层嵌套编码机制的混合差分进化的入侵杂草调度算法,该算法引入具有学习机制的算子以改善解的质量.为有效提高算法的收敛性,在变异及邻域操作中考虑自适应过程.仿真实验结果表明,该算法是有效且可行的,优化晶舟分配的调度较未优化的调度可提高至少10%的性能.%To efficiently address the multi-order-per-job ( MOPJ) scheduling problem of a single machine with multiple preprocess constraints in wafer fabrications, including setup time, uncertain allocation of front opening unified pods ( FOUPs ) , machine parameter adjustment, a scheduling problem domain was described and a mathematical programming model was set up with an objective of minimizing total completion time, and several theorems were established to obtain superior feasible solutions, in addition, a hybrid invasive weed optimization algorithm combined with differential evolution and adopted a two-level encoding mechanism was developed, in which the learning mechanism was introduced to enhance the quality of the solution. Moreover, adaptive process was applied to the mutation and neighborhood search to effectively improve the algorithm convergence. Finally, simulation results verify the validness and feasibility of the proposed algorithm and show that a 10% improvement is made on the performance by the scheduling approach.
展开▼
