Exponentially weighted moving average (EWMA) 回饋控制器為半導體科技產業中常被使用的製程批次控制手法。現存文獻大多針對單一機台生產單一產品的生產型態下，探討控制器之績效表現及分析。然而目前國內積體電路 (IC)製程大多採用多產品的混合生產型態 (mixed-run production)。針對此類混合生產型態，文獻上雖已有學者提出 tool-based (TB)、 product-based (PB) 、 cycle resetting (CR) 及 modified tool-based (MTB) 等回饋控制器，但上述控制器皆未考慮具有線性漂移 (linear drift)的製程，致使當製程發生線性漂移現象時，製程產出值的 total sum square error (TSSE) 有過大之現象，因此績效並不理想。針對具有線性漂移的製程，已有論文提出 modified tool-based of drifted-version (DMTB) 回饋控制器來處理上述問題，並能有效地降低切換產品後前幾期之製程產出值與目標值的偏差量，但是在連續相同產品製程中，製程產出期望值並不會收斂在目標值上，針對此項不足，本文建構出 modified tool-based with double EWMA (簡稱 MTB-dEWMA) 控制器來改善這個問題。本文推導出此新控制器製程產品切換後第一期產出公式及穩定條件，並以模擬實驗探討 MTB-dEWMA 與 DMTB 、 MTB 及 CR 控制器之 TSSE 績效分析及偏差問題，結果顯示 MTB-dEWMA 控制器的表現優於 DMTB 、 MTB 及 CR 控制器。

Exponentially weighted moving average (EWMA) feedback controller has been widely used in semiconductor manufacturing processes. The long-term stability and short-term performance of EWMA control scheme have been addressed by several studies in the literatures. Most of the controllers emphasize on the case of a single-tool with single-product production situation. However, in advanced manufacturing, multiple-product production situations are common for the practical implementation of run-to-run (R2R) control scheme. Recently, tool-based (TB), product-based (PB), cycle resetting (CR) and modified tool-based (MTB) controllers have been proposed to handle the mixed-run R2R control problem. These controllers are developed under the scenarios of non-drifted processes. For drifted-processes, these control schemes may suffer from a larger total sum square error (TSSE). The drifted-version MTB (DMTB) controller partially solves this problem, leading to a smaller TSSE (of output) but still having a non-ignorable bias during the same product production run. To overcome this issue, this thesis propose a modified DMTB with an extra controller to compensate for drift, called MTB-dEWMA. The stability conditions of MTB-dEWMA controller under the assumption that the process disturbance follows general time series are discussed. Furthermore, a simulation is conducted to examine the performance of MTB-dEWMA controller with existing competitors, including DMTB, MTB and CR. The results show that MTB-dEWMA controller outperforms other competitors in terms of TSSE criterion.

[1] Butler, S.W. and Stefani, J.A. (1994). “Supervisory run-to-run control of a polysilicon gate etch using in situ ellipsometry,” IEEE Transactions on Semiconductor Manufacturing, 7(2), 193–201.
[2] Del Castillo, E. (1999). “Long run and transient analysis of a double EWMA feedback controller,” IIE Transactions, 31(12), 1157-1169.
[3] Del Castillo, E. (2001a). “Some properties of EWMA feedback quality adjustment schemes for drifting disturbances,” Journal of Quality Technology, 33(2), 153-166.
[4] Ingolfsson, A. and Sachs, E. (1993). “Stability and sensitivity of an EWMA controller,” Journal of Quality Technology, 25, 271-287.
[5] Sachs, E., Guo, R. S., Ha, S. and Hu, A. (1991). “Process control system for VLSI fabrication,” IEEE Transactions on Semiconductor Manufacturing, 4(2), 134-144.
[6] Tseng, S. T., Chou, R. J. and Lee, S. P. (2002). “Statistical design of double EWMA controller,” Applied Stochastic Model in Business and Industry, 18, 633-643.
[7] Zheng, Y., Ai, B., Wong, D.S.H., Jang, S. S., Wang, Y., Zhang, J. (2010). “An EWMA Algorithm With a Cycled Resetting (CR) Discount Factor for Drift and Fault of High-Mix Run-to-Run control,” IEEE Transactions on Industrial Informatics, 6, 229-242.
[8] Zheng, Y., Lin, Q.H., Wong, D.S.H., Jang, S. S., Hui, K. (2006). “Stability and performance analysis of mixed product run-to-run control,” Journal of Process Control, 16, 431-443.
[9] 李家豪 (2012)，《多產品混合生產製程之批次回饋控制分析》，國立清華大學統計學研究所碩士論文。
[10] 葉佳和 (2013)，《機台老化之多產品混合生產的批次回饋控制分析》，國立清華大學統計學研究所碩士論文。
[11] 簡勁舟 (2011)，《單機台多產品批次回饋控制之績效分析》，國立清華大學統計學研究所碩士論文。