一种基于随机规划的晶圆制造产能规划方法

Stochastic-Programming-Based Capacity Planning for Wafer Fabrication

针对晶圆制造过程中产能与需求的高度不确定性问题,采用基于场景(scenario)的随机规划方法进行设备购置决策.未来不确定的需求采用多个以发生概率相联系的预测场景来模拟,使得到的设备购置决策具有较好的鲁棒性.采用加工作为衡量产能的标准,考虑了优先设备与后备设备的区别,目标是尽量满足各阶段各场景中的需求.结果表明,当未来产品需求在一定范围内变动时,基于随机规划的决策优于基于线性规划的决策.

Because of the high uncertain capacity and demand, capacity planning is very difficult in wafer manufacturing. To cope with this problem, a stochastic-programming-based method was used to make tool procurement decision, which is robust to change in demand. It considers several discrete demand scenarios with associated probabilities to determine the tools to buy under certain budget constraint. The number of operations is taken as the method of measuring capacity; and the difference between primary and backup machine is considered; and the objective is to meet the demand in every scenarios. It turns out that when the demands fluctuate within a certain range, the stochastic-programming-based decision is better than the linear-programming-based decision.