A focus on cycle time-vs.-tool utilization "paradox" with material handling methodology

From large scale factory simulation, the semiconductor manufacturing process is characterized by a fixed nonlinear relationship between mean factory cycle time and overall average tool utilization. This allows selection of one parameter, which thus determines the other, for a specific factory. Although on a local level, the relationship between cycle time and tool usage is predicted via queuing theory, we find similar results for the entire large scale factory. On this scale, the relationship is paradoxical as it does not allow short mean cycle times and high average tool usage concurrently. Net improvements in factory performance can be had only by moving this relationship to a parallel curve, more favorable in the cycle time vs. tool utilization domain, which requires fundamental system changes. Emphasis on such shifts currently focuses on methodologies locally synchronizing asset use. This paper, however, considers the aggregate factory, where local asset usage is determined only by the random influence of the factory. Local assets are considered to have ideal efficiency, and thus only the stochastic factory relations determine their performance. Such external influences are WIP availability, WIP denomination and WIP rate. With this global factory view, WIP handling methodologies have a first order effect on overall performance, which promote migration from one factory characteristic curve to a higher performance one purely through material handling changes. Dynamic discrete event simulation, with 0.5 s resolution of WIP tracking, is used to find first order effects of transport methodologies on the total system.