以MPLS的模式結構進行最後產品品質之批次間與本批次間的控制

Batch-to-Batch and Within-Batch Control Design for Final Product Quality Under MPLS Model Structure

林昆棋(Kuen-Chi Lin)

指導教授 ： 陳榮輝

中原大學/工學院/化學工程學系/碩士(2006年)

https://doi.org/10.6840/cycu200600625

摘要

本文主要針對批次出口品質規格提出兩套設計最適操作條件的控制策略，其一為一階段MPLS控制策略，另一則是二階段MPLS控制策略。而主要的設計概念則是以MPLS(Multi-way Partial Least Squares)為基礎，輔以雙指數加權移動平均(Double Exponentially Weighted Moving Average, dEWMA)修正模式偏差。就一階段MPLS控制策略而言，我們主要利用操作變數及量測變數對品質變數建立一經驗模式；而二階段MPLS控制策略除利用操作變數與量測變數建立一經驗模式外，並利用量測變數與品質變數建立另一以MPLS為基礎的經驗模式。

由於批次製程產品的多變性與多樣性，要於短時間內對製程推導物理模式以架構控制策略實屬不易。因此，我們利用以MPLS為基礎之經驗模式來處理此類問題。此模式可有效利用批次操作數據並萃取其間輸入與輸出變數最強關聯性，進而有效濾除雜訊、去除變數與變數之間交互關聯性。而為避免模式失真或未知擾動影響，我們將MPLS架構拆解為子空間各個獨立單環路，並導入dEWMA控制器以疊代學習修正模式偏差與衰退問題。

除此兩套控制策略外，我們亦導入傳統批次控制設計的批次間與本批間控制概念。在批次間控制方面，主要利用批次生產完畢所量測之品質變數經回饋控制修正模式偏差，並以之計算下一批次最佳操作配方。而為消除批間干擾，我們亦將本批間控制概念導入前述兩套控制策略中，利用線上量測資訊即時修正後續操作軌跡，以使批次生產完畢之產品品質符合規格。

在本文中，也將以一批次反應製程為例，分別針對設定點改變製程、批間干擾製程與批次衰退測試製程測試本文所提方法之控制效能。並與近期相關文獻方法作比對，驗證我們所提方法於真實機台操作的優勢。

關鍵字

批間控制 ； 最小平方法 ； 批次控制 ； 品質控制

並列摘要

The two novel design strategies, one-stage MPLS based dEWMA control and two-stage MPLS based dEWMA control, respectively, are proposed for maintaining final product properties. They are model-based control algorithms using multi-way partial least squares (MPLS) models and double exponentially weighted moving average (dEWMA). In one-stage MPLS based dEWMA control, only one MPLS model is used to relate the final quality with the on-line measured and the manipulated variables. On the other hand, in two-stage MPLS based dEWMA control, one MPLS model is used to relate the final quality with the on-line measured variables while the other can relate the on-line measured variables with the manipulated variables. The MPLS based model utilizes input/output trajectory information as historical databases. It can address the difficulties in developing detailed mechanistic models. Rather than dealing with the high dimensional manipulated variable trajectories in the real space, MPLS takes the advantage of extracting the strongest relationship between the input and the output variables in the reduced space of the latent variables model. It is particularly useful for inherent noise suppression and removal of interaction of variables. Also, it can break the MIMO process into a series of univariate processes. However, MPLS may have the problems of the model-plant mismatches, so the dEWMA control algorithm is applied separately and directly to each SISO control loop based on model prediction errors from the previous batch run.

With the updated process model, the feedback control of the output quality is introduced to find out the adjusted manipulated variables and track the desired quality at the end-point of a batch run. The batch-to-batch scheme can not only reduce the model mismatch but also make the end point quality gradually reach the desired quality. Furthermore, in order to adjust the operation in time, the MPLS based dEWMA control strategy is developed to explore the possible adjustments of the future input trajectories. It fixes up the disturbances just in time instead of until the next batch run and keep the product within specifications when this batch is finished.

The effectiveness of the proposed design strategy is demonstrated through three cases of a simulated batch process, including setpoint changes, disturbance changes within a batch run, and disturbance changes occurred at batch-to-batch. These cases are used to investigate the potential applications of the proposed method and make a comparison with some conventional methods recently reported in the literature.

並列關鍵字

Quality Control ； Within-Batch Control ； Batch-to-Batch control ； Multiway partical least squares

參考文獻

1.Bakshi, B.R. and Stephanopoulos, G., “Representation of process trends - IV. Induction of real-time patterns from operating data for diagnosis and supervisory control,” Comput. Chem. Eng., 18(4), 303 (1994)

3.Butler, S. W. and Stefani, J. A., “Supervisory Run-to-Run Control of a Polysilicon Gate Etch Using in Situ Ellipsometry,” IEEE Trans. Semicond. Manuf., 7, 193 (1994)

4.Chen, J. and Cheng, Y. C., “Applying PLS-Based Decomposition Structure to Multi-Loop Adaptive PID Controllers in Nonlinear Processes,” Ind. Eng. Chem. Res., 43(18), 5888 (2004)

5.Crowley, T. J.; Choi, K. Y., “Experimental studies on optimal molecular weight control in a batch-free radical polymerization process,” Chem. Eng. Sci., 15, 2769 (1998)

6.Del Castillo, E. and Hurwitz, A., “Run to Run Process Control: Literature Review and Extensions,” J. Qual. Technol., 29, 184 (1997)