Using simulation modeling to calculate WIP levels in semiconductor manufacturing

Summary form only given, as follows. This paper discusses a modeling approach which is used to support the production control methodology called AMD Kanban System. AMD Kanban System is a simple, visual, self-regulating system for controlling inventory levels at operations within a manufacturing process. These visual signals pull inventory as required from prior steps, thus maintaining line balance while meeting production schedules with minimum work in process. This methodology involves determining the WIP levels at each step or at the beginning of the "linked" steps in the process flow and adjusting them on a periodic basis. Calculating the exact target inventory levels is a challenging task. The inventory levels are currently estimated based on the processing steps included in the link. The methodology presented in this paper addresses the issue of calculating these inventory levels using simulation modeling. A simulation model of three major semiconductor process flows is used as a test bed. The model is adaptable for any time period the user chooses, e.g. a month or a full quarter. The model predicts the bottlenecks in the fab accurately and allows ideal inventory levels at each step in the process flow assigned by the user. The software then reports exactly what quantity of the inventory assigned each step was used during the simulation period. Based on this it is possible to predict the inventory needed at each step. A prototype implementation method is presented.