AbstmctThis paper descrlbes research on the development of a high-fidelity simulator using the object-oriented programming paradigm.The simulator models the complex interactions in reentrant flow manufacturing facilities. In addition to potentially providing better estimates on throughput, work-ln-process levels, and bottleneck resources in the modeled system, the simulator serves as a platform to investigate h u e s of operator cross-training, decision aiding, and the effect of operator decisions in such manufacturing facilities. We describe the class hierarchy of objects required for modeling and analysis of semiconductor and thin films fabrication and discuss our approach to modeling complex supervisory control decisions in re-entrant flow lines.
I. BACKGROUNDSemiconductor manufacturing is a complex process which can be broadly viewed in four stages: raw wafer manufacturing, wafer fabrication, probe and test, and packaging [ 11. Our work focuses on the fabrication process. Semiconductor fabrication involves the processing of silicon wafer lots by a re-entrant flow process. Several hundred processing steps are required to buildup integrated circuits (IC) layer by layer. IC fabrication facilities cost several hundred million dollars due to requirements for cleanliness, vibration control, and chemical punty specified at the submicron level [5,6]. Such a re-entrant flow also characterizes thin films fabrication, as it applies to high performance circuit packaging. Modeling and analysis of the IC fabrication lines, both at the design and operational stages, are critical in achieving enhanced productivity required in these capital-intensive industries. Queueing network models and simulation models are widely used. both by the academic and industrial research communities [l, 2.4.91. Although a variety of modeling techniques are useful in the design and performance analyses of these facilities, simulation remains the most flexible. Complexities in fabrication facilities can be modeled effectively using the simulation methodology [l, 91.Most current simulation models do not have the high level of fidelity necessary to model the dctailed interactions in the fabrication plant. For instance, it is difficult to model complex line ' control decisions (e.g.. set-up avoidance), operator skill levels and decisions (e.g., rework judgements and send-aheads), and additional tool constraints (e.g., preventive maintenance). This leads to an incomplete and possibly incorrect model, potentially limiting the value of the modeling effort. Despite recent advances in simulation software, two major problems in applying the simulation methodology are model development and model validation. The one to one correspondence between objects in the real problem and their abstractions in the software provided by the OOP paradigm offer significant potential for better modeling [7. lo]. We have used a combination of principles from control theory, modular systems development theory, and the OOP paradigm in developing a high fidelity simulator to sup...
