| 520 |
|
|
|a Semiconductor manufacturing is a high-technology industry which is capital intensive and operationally complex with its process technology refreshed every two years. Precision in capacity planning is critical to ensure the right amount of capital equipment is purchased to match the demand while meeting aggressive cost and operational targets. The key input parameter for capacity calculations is the equipment output rate. As equipment get more complex, its output rate become difficult to predict using spreadsheets, thus the need for detailed dynamic equipment simulation models. However, literature on how to build detailed equipment simulation models for real-world is scarce. Practitioners do not share their experience openly due to proprietary reasons. This dissertation investigates the complexity of semiconductor manufacturing which makes its capacity planning difficult. The techniques to build, verify and validate high fidelity equipment simulation models were developed. The models are then used to augment capacity planning and productivity improvement decision making. Case studies are conducted using the models to improve capacity forecast planning accuracy for capital purchase decisions which resulted in million dollars capital avoidance, test equipment productivity improvement ideas and decide which ones have benefits to pursue, and determine the effect of different operator manning ratios for manufacturing execution decisions. The results show that raw model accuracy can be up to 99% using the methods described here. For manufacturing execution, model accuracy can be up to 95% due to variability in human performance, but good enough to provide insights on manning ratio strategies. The case studies demonstrate how the results directly contribute to company performance in terms of capital efficiency, capital expenditure avoidance, and waste reduction. It enables optimal equipment configuration decisions to be made upfront during technology development. It also earns credibility and senior management confidence in using such simulation models for decision making.
|
