In semiconductor manufacturing discrete event simulation systems are quite established to support multiple planning decisions. During the recent years, the productivity is increasing by using simulation methods. The motivation for this thesis is to use online simulation not only for planning decisions, but also for a wide range of operational decisions. Therefore an integrated online simulation system for short term forecasting has been developed. The production environment is a mature high mix logic wafer fab. It has been selected because of its vast potential for performance improvement. In this thesis several aspects of online simulation will be addressed: The first aspect is the implementation of an online simulation system in semiconductor manufacturing. The general problem is to achieve a high speed, a high level of detail, and a high forecast accuracy. To resolve these problems, an online simulation system has been created. The simulation model has a high level of detail. It is created automatically from underling fab data. To create such a simulation model from fab data, additional problems related to the underlying data arise. The major parts are the data access, the data integration, and the data quality. These problems have been solved by using an integrated data model with several data extraction, data transformation, and data cleaning steps. The second aspect is related to the accuracy of online simulation. The overall problem is to increase the forecast horizon, increase the level of detail of the forecast and reduce the forecast error. To provide useful forecast results, the simulation model contains a high level of modeling details and a proper initialization. The influences on the forecast quality will be analyzed. The results show that the simulation forecast accuracy achieves good quality to predict future fab performance. The last aspect is to find ways to use simulation forecast results to improve the fab performance. Numerous applications have been identified. For each application a description is available. It contains the requirements of such a forecast, the decision variables, and background information. An application example shows, where a performance problem exists and how online simulation is able to resolve it. To further enhance the real time capability of online simulation, a major part is to investigate new ways to connect the simulation model with the wafer fab. For fab driven simulation, the simulation model and the real wafer fab run concurrently. The wafer fab provides several events to update the simulation during runtime. So the model is always synchronized with the real fab. It becomes possible to start a simulation run in real time. There is no further delay for data extraction, data transformation and model creation. A prototype for a single work center has been implemented to show the feasibility.    «

In semiconductor manufacturing discrete event simulation systems are quite established to support multiple planning decisions. During the recent years, the productivity is increasing by using simulation methods. The motivation for this thesis is to use online simulation not only for planning decisions, but also for a wide range of operational decisions. Therefore an integrated online simulation system for short term forecasting has been developed. The production environment is a mature high mix...    »