Effective WIP flow estimation for daily fab production target setting with consideration of variability

This paper presents two innovative designs for including operation variability into an existing wafer flow estimation scheme SOPEA for the purpose of daily fab production target setting by processing stage and by product. Two types of fab operation variability are considered: (i) the machine allocation variability of individual stages caused by dynamic allocation of machines to individual stages, and (ii) the workload variability caused by the compound effect of uncertain processing time, initial WIP distribution and wafer start of the day under given machine allocation. An ON/OFF model with parameters fit to empirical data is proposed to capture the wafer flow slow down by machine allocation variability. Probabilistic queueing analysis is then performed to compensate the workload variability. The two methods have also been integrated with the DTS system in a memory fab case. With the consideration of machine allocation variability, SOPEA overestimation is reduced by 25% and 6% in average closer to actual daily flow-ins than the empirical rules.