An easy approach to extending a short term simulation model for long term forecast in semiconductor industry

Abstract

The operational decision making in the BOSCH's 200mm wafer fabrication facility has been guided by short term simulation forecasts. The forecasts provides the capability of identifying daily bottlenecks, forecasting daily fab outs, optimizing the preventive maintenance plans and personal resource planning. Now there is a pressing need to extend the forecast time horizon to several months for making decisions such as analyzing different ramp up scenarios, evaluating the impact of dispatch rules, identifying bottlenecks for capital investment, etc. As the short term model has achieved forecast accuracy of above 90%, it is used as the basis to generate the long term model. In this paper, we discuss the key issues associated with this model generation process. These issues are: process flows compression, flexible equipment dedications, model warm-up, wafer start generation, and future fab capacity changes. Our approach enables us to use the same model generation framework for both models.