We study the performance of production planning models for wafer fabs in a rolling horizon setting under stochastic demand modeled using the Martingale Model of Forecast Evolution (MMFE). The models differ in how they incorporate lead times and safety stocks. The number of frozen periods is considered as an experimental factor to study the nervousness of the resulting plans. Our simulation experiments show that chance-constrained production planning models that integrate safety stock and production planning decisions outperform others. The formulations with workload-dependent lead times also outperform those with fixed lead times that are an integer multiple of the period length.

中文翻译：

---

半导体晶圆制造中具有安全库存和预测演变的生产计划模型的基于仿真的性能评估

我们研究了在使用 Martingale 预测演化模型 (MMFE) 建模的随机需求下滚动范围设置中晶圆厂生产计划模型的性能。这些模型的不同之处在于它们如何结合交货时间和安全库存。冻结期的数量被认为是研究结果计划紧张程度的实验因素。我们的模拟实验表明，将安全库存和生产计划决策相结合的机会约束生产计划模型优于其他模型。工作负载相关提前期的公式也优于周期长度整数倍的固定提前期的公式。