Multi-objective capacity planning for agile semiconductor manufacturing

Abstract

Recent semiconductor business requires an agile fabrication facility (fab) that produces wafers in short cycle time and at low production cost to promote customer satisfaction. In this paper, a multi-objective capacity planning methodology is considered. First, a semiconductor wafer fab is modelled as an open queueing network in which each tool group is a node with a GI/G/m queue. Then, cycle time and production cost are modelled as functions of the number of tools of each tool group and of the throughput of each product. This enables to propose a capacity planning which has a property to quantify the trade-off between cycle time and production cost. The proposed capacity planning can be used to determine a tool-set configuration that satisfies required levels of cycle time and production cost with budget constraint, and to quantify the effect of fab scale on cycle time and production cost. The proposed capacity planning is also applied to the SEMATECH data to show how the following phenomena can be measured. (1) The larger the fab scale, the shorter the cycle time and the lower the production cost. (2) One-machine tool group causes higher production cost or longer waiting time in minifabs. (3) Versatile tools decrease the minimum fab scale that satisfies the required cycle time and production cost.