批次控制在半導體生產製程中扮演極重要的監控角色。既有文獻之批次回饋控制方法大多建構在SISO系統且製程之I-O模型為靜態模型下，然而，生產線實際上卻為MIMO系統且製程同時具有遷移效應較為常見；文獻上有qMMSE控制器被提出來監控此類動態生產製程，由於此控制器僅限於當製程I-O模型為 TF (1,1,1) + VIMA (1,1) 下被建構出來；因此當模型為一般化的 TF (r,s,1) + VARIMA (p,d,q) 模型時，其製程的TMSE之表現並不理想。針對此缺點，本論文提出修正的一般化動態qMMSE回饋控制器，並分別推導出此控制器底下的製程產出公式、製程穩定條件，並同時針對既有的文獻進行理論修正。最後，本文亦探討當實際製程I-O模型為 TF (2,2,1) + VARIMA (1,1,1) 時，此真實模型底下的控制器與模型誤判後的控制器之績效表現；結果顯示，本文提出的控制器之優點為不僅能將製程產出值更快速地調整到目標值，同時亦可使製程的TMSE大幅降低，也能夠發現製程干擾項的誤判對於控制器將會產生嚴重的影響。

Run-to-run control plays a critical role in monitoring or adjusting the IC manufacturing process. Traditionally, in literature, most of the process I-O model deals with single-input single-output (SISO) static model. However, it’s more common to see that the process I-O model follows the multiple-input multiple-output (MIMO) with a dynamic carry-over effect model. Recently, a qMMSE controller has been proposed to monitor such a dynamic MIMO process. The controller, however, is restricted to the case that the process I-O model follows TF (1,1,1) with VIMA (1,1) disturbance. This qMMSE controller, in general, does not perform well when the I-O model follows TF (r,s,1) with VARIMA (p,d,q) disturbance. To overcome this difficulty, this study first constructs a generalized MIMO dynamic qMMSE controller. Then, the process output formula and the stability conditions of the proposed controller are derived, respectively. We use a CMP example to illustrate the proposed method. The results demonstrate that the proposed controller not only adjusts the process output value more quickly to the target value, but also reduces the TMSE of the process significantly. Finally, when the actual I-O model follows TF (2,2,1) + VARIMA (1,1,1), this study uses a simulation study to address the misspecification effects of using wrong controllers. The simulation result shows that using a wrong controller will have serious impacts on the process stability conditions and performance of process TMSE.

摘要
第一章 簡介 1
1.1 前言 1
1.2 批次控制介紹 2
1.3 研究動機 5
1.4 研究架構 7
第二章 文獻回顧及問題描述 8
2.1文獻回顧 8
2.2前述qMMSE控制器的M矩陣之修正 9
2.3 問題描述 10
第三章 一般化動態qMMSE回饋控制器分析 11
3.1 一般化動態qMMSE回饋控制器之建構 11
3.2 一般化動態qMMSE回饋控制器控制後產出迭代公式 12
3.3 一般化動態qMMSE回饋控制器的穩定條件 13
3.4 小結 16
第四章 控制器實例績效分析 17
4.1 一般化動態qMMSE回饋控制器在CMP製程之應用 17
4.2 當預測模型發生誤判時，對控制器績效之影響 20
4.3 當預測模型發生誤判時，影響控制器績效之模擬分析 27
第五章 結論與後續研究 30
附錄 32
參考文獻 56

[1] 米忻超 (2017). “動態系統暨多產品混合生產製程之多變量批次控制器研究,” 國立清華大學統計學研究所博士論文.
[2] 陳沛瑜 (2015). “批次動態回饋系統的一般化qMMSE控制器分析,” 國立清華大學統計學研究所碩士論文.
[3] 吳若華 (2017). “動態Quasi MMSE回饋控制器之建構與績效分析,” 國立清華大學統計學研究所碩士論文.
[4] Box, G. E. P., Jenkins, G. M., and Reinsel, G. C. (1994). Time series analysis: forecasting and control, 3rd ed. Prentice Hall, Englewood Clifs, NJ.
[5] Butler, S. W. and Stefani, J. A. (1994). Supervisory run-to-run control of polysilicon gate etch using in situ ellipsometry. Semiconductor Manufacturing, IEEE Transactions on Semiconductor Manufacturing, 7(2), 193-201.
[6] Capilla, C., Ferrer, A., Romero, R., and Hualda, A. (1999). Integration of statistical and engineering process control in a continuous polymerization process, Technometrics, 41, 14-28.
[7] Chen, A., and Guo, R. S. (2001). Age-Based double EWMA controller and its application to CMP processes, IEEE Transactions on Semiconductor Manufacturing, 14, 11-19.
[8] Del Castillo, E. (1999). Long run and transient analysis of a Double EWMA feedback controller, IIE Transactions, 31, 1157-1169.
[9] Del Castillo, E. (2001). Some properties of EWMA feedback quality adjustment schemes for drifting disturbances. Journal of Quality Technology, 33, 153-166.
[10] Del Castillo, E. (2002). Statistical process adjustment for quality control, Wiley Interscience.
[11] Fan, S. K. S., Jiang B. C., Jen, C. H., and Wang, C. C. (2002). SISO run-to-run feedback controller using triple EWMA smoothing for semiconductor manufacturing process. International Journal of Production Research, 40, 3093-3120.
[12] Ingolfsson, A. and Sachs, E. (1993). Stability and sensitivity of an EWMA controller. Journal of Quality Technology, 25, 271-287.
[13] Jen, C. H. and Jiang, B. C. (2008). Combining on-line experiment and process control methods for changes in a dynamic model. International Journal of Production Research, 46, 3665-3682.
[14] Reinsel, G. C. (2003). Elements of multivariate time series analysis, New York, NY: Springer Science & Business Media.
[15] Tsay, R. S. (2013). Multivariate time series analysis: with R and financial applications. John Wiley & Sons.
[16] Sachs, E., Hu, A., and Ingolfsson, A. (1995). Run by run process control: combining SPC and feedback control, IEEE Transactions on Semiconductor Manufacturing, 8, 26-43.
[17] Tseng, S. T., and Chen, P. Y. (2017). A generalized quasi-MMSE controller for run-to-run dynamic models. Technometrics, 1-10.
[18] Tseng, S. T., and Lin, C. H. (2009). Stability analysis of single EWMA controller under dynamic models. IIE Transactions, 41(7), 654-663.
[19] Tseng, S. T., and Mi, H. C. (2014). Quasi-minimum mean square error run-to-run controller for dynamic models. IEE Transactions, 46(2), 185-196.
[20] Tseng, S. T., Mi, H. C., and Li, I. C. (2016). A multivariate EWMA controller for linear dynamic processes, Technometrics, 58(1), 104-115.