提出了一种滚动式基于场景(scenario)的随机规划方法.由于需求的高度不确定性,采用多个预测场景来模拟需求的不确定性,其发生概率之和为1;由于一个产能计划周期可能包含多个需求预测周期,以分别满足几个预测周期的需求为约束,而不是满足整个计划周期的需求;以滚动的形式来做设备购置计划,即根据未来几个计划周期的情况来做本期设备购置决策,使得到的设备购置决策具有更好的鲁棒性.结果表明,当未来产品需求在一定范围内变动时,基于随机规划的滚动计划决策优于原来的随机规划决策.
Because of high uncertain capacity and demand, the capacity planning for wafer fabrication is very difficult. To cope with this problem, a rolling planning method was used based on an improved scenario-based-stochastic-programming to make tool procurement decision. The uncertainty of product demand is represented with a set of demand scenarios with associated probabilities of occurrence. One capacity planning period may contain several forecasting periods. The demands of forecasting periods, rather than the planning period, are considered in the constraints. The tool procurement plan is made in sequence by considering the status of the next several planning periods. All of those make the tools procurement robust. It turns out that the decisions based on the rolling planning method which considers the planning periods and forecasting periods are better than the original stochastic-programming-based decision.