考虑设备时变效应的晶圆加工序列决策调度方法

为改善半导体生产过程中设备状态不确定引起的时变效应可能造成生产计划难以推进、生产效率下降等问题，使用考虑设备时变效应的晶圆加工序列决策调度方法制定调度方案。采集过往加工工时数据，挖掘设备状态变化的特征参数与晶圆的加工工时时变效应的关联关系，从而建立考虑时变效应的平行机调度模型，实现最大完工时间的最小化。设计集成调度优化知识的混合搜索算法(HSAOSK),利用单机调度最优规则与多机调度优化知识库减少搜索空间，提高算法的计算效率。实际算例的分析结果表明：HSAOSK算法求解小规模算例的最优解与精确算法(BRA)相同，求解大规模算法时与其他优化算法相比，最大完工时间可减少6.17%,且计算时间非常短，HASOSK算法的优越性能满足构建半导体调度决策方案的需求。调度决策方法不仅能为具有时变效应的半导体生产系统提供有效的加工序列决策，还能针对设备状态提供不同的维护决策以保证生产效率。

Wafer production sequence scheduling considering time-changing effects

To solve the problem of impediments of production plans and decline of production efficiency caused by potential time-varying effects of the machine condition uncertainty in the semiconductor production process, decision-making methods for wafer production sequence scheduling considering time-changing effects are developed. Firstly, with collecting historical processing time data, the relevance between characteristics of machine condition variation and time-changing effects of wafer processing times is diagnosed to establish parallel machine scheduling model considering time-changing effects. The target is to minimize the makespan. A hybrid search algorithm with optimal scheduling knowledge (HSAOSK) is designed based on optimal single machine scheduling rules and multi-machine scheduling optimization knowledge to reduce the searching space and improve the calculation efficiency. Computational experiments show that the optimal solution of the HSAOSK algorithm is the same as the exact algorithm to solve small-scale cases. As for the large scale cases, comparing to the other algorithms, the optimal makespan of HSAOSK algorithm has 6.17% decrement with the shortest time consumption. The HSAOSK algorithm can meet the needs of constructing a semiconductor scheduling decision-making scheme.