The typical approach taken to simulating semiconductor wafer fabs is a timeand memory-intensive process that tracks individual wafers as they pass through the fab. This modeling paradigm is known as a job-driven simulation. In this work, we describe research done at UC Berkeley in cooperation with Intel, the Semiconductor Research Consortium, and SEMATECH to develop a resource-driven simulation to modeling fabs. Developing the new model provided interesting insights into modeling assumptions typically made. It also showed that some aspects of the fab can be modeled more accurately in the resource-driven model, while others are more accurately modeled using a job-driven approach. However, the resource-driven model runs considerably faster, making more extensive experimentation possible.
[1]
Bryan L. Deuermeyer,et al.
Deadlock detection and resolution for discrete-event simulation: multiple-unit seizes
,
1998
.
[2]
Lee W. Schruben.
Graphical Simulation Modeling and Analysis: Using Sigma for Windows (The Scientific Press Series)
,
1995
.
[3]
Lee W. Schruben,et al.
Resource graphs for modeling large-scale, highly congested systems
,
2001,
Proceeding of the 2001 Winter Simulation Conference (Cat. No.01CH37304).
[4]
Lee W. Schruben,et al.
Mathematical programming models of discrete event system dynamics
,
2000,
2000 Winter Simulation Conference Proceedings (Cat. No.00CH37165).