批次控制已廣泛應用於半導體製造。然而,在一般批次控制中,並沒有任何機制來檢測製程模型可能發生的變化。在本研究中,針對兩種生產型態進行了研究。一個是針對多產品和多機台的生產方式,整合了製程監測及反饋控制。運用管制圖來檢測製程模型的變化應歸屬於產品或機台。管制圖除了用來監測製程模型的變化,也用來計算如何調整製程模型。另一個是Partial Least Squares (PLS)的雙加權移動平均控制器(D - EWMA)應用於多輸入多輸出(MIMO)製程。它監測製程模型,當模型發生變化時,適時修正製程模型。當製程模型改變時,它的一階差分潛在控制作用將更為有效地檢測模型的變化。不論是在系統模擬,或在工業界的實例應用,均証實了本方法的優越性。
Run to run control scheme has been widely used in semiconductor manufacturing process. However, in conventional run-to-run control, there is no mechanism to detect possible changes in the process model. In this study, two production type have been studied. One is a novel run-to-run controller that incorporates process monitoring into feedback control is proposed for a multiple product and multiple tool process. A control chart is designed to detect changes of process gain of the linear input-output relationship attributable to individual products or tools. The information contained in the control chart is used to adapt the process changes. Another application is a partial least square (PLS) based double exponentially weighted moving average controller (d-EWMA) of multiple-input and multiple-output (MIMO) processes subjected to constant or moderate gain shifts disturbance. It is mainly focused on a new gain monitoring method and a PLS based model updating. It was demonstrated that under the circumstance the variance of the first order difference of the latent control action would be more effective in detecting the gain shifts. Effectiveness of this method was demonstrated using simulations as well as a real industrial application.