| Summary: | 碩士 === 銘傳大學 === 應用統計資訊學系碩士班 === 98 === For the last two decades, the run-to-run controller has become very popular to enhance the performance of semiconductor manufacturing processes by integrating feedback control and statistical process control. Due to lot-to-lot feedback adjustment, the recipe of the process is varied in lot. It induces that process output is not only auto-correlated, but also has variation in terms of lot, wafer and the sites of a wafer.
In this study, according to the framework of run-to-run control for semiconductor manufacturing process, we use the state space model to describe the process output by Single Exponentially Weighted Moving Average (SEWMA) feedback. According the proposed model, the state prediction controller (STPC) was proposed to monitoring the expected means for each lot.
Based on the assumption that the process output adjusted by the feedback controller is stationary, we use the criteria of average run length to evaluate the performance of our proposed STPC. Compare to the traditional Shewhart control chart, we have the two main results as follows:
(1) The ARL0 of our proposed control chart is very close to 370.4.
(2) When the process is shift, the proposed control chart can detect the assignable cause quickly.
|
|---|
