批次剖面資料之錯誤偵測分析

Fault Detection Analysis of Batch Profile Data

趙安國(An-Kuo Chao)

指導教授 ： 曾勝滄

國立清華大學/理學院/統計學研究所/碩士(2008年)

https://doi.org/10.6843/NTHU.2008.00589

摘要

批次製程目前已廣泛應用於化工、醫藥、半導體等製程工業，因為產品品質特徵值不容易及時取得，造成執行線上製程管制的困難，所以如何有效利用製程資料，執行製程的錯誤偵測與分類，是目前製程管制的重要研究課題。製程資料在批次之間的變異可以分為時間變異和變數變異，這兩種變異會影響製程管制的準確度而且通常與產品品質有關，所以必須以尺度參數表示，從製程資料中抽取出來。本文對批次剖面資料提出一套統計分析流程，先將批次剖面資料分為時間量度、位移量度和抽離位移量度後的殘差三部分，分別表示批次之間製程時間的差異、變數的平移或漂移量和沒有平移或漂移現象的批次剖面資料。然後用改善的健康指標分析殘差，並對分離出的尺度參數提供監控與分析的方法，除了能夠完備的監控製程中是否存在異常現象，亦可以研究製程變數之間的關係，以達成執行製程管制與瞭解並改善製程之目的。

關鍵字

批次製程 ； 錯誤的偵測與分類 ； 資料採礦 ； 健康指標 ； HASH(0x1cd63090)

並列摘要

Batch process has been widely applied in the chemicals, pharmaceutical, and semiconductor industries. Based on the limitation of measurement devices, we cannot measure the quality characteristics of products immediately and then have no information to monitor and control. Hence, how to use the fault detection and classification (FDC) methodology to find the root cause by using data of the process variables is an important issue. In general, the variability of process variables between batches can be divided into the time axis variability and the variable axis variability. Both of them would affect the preciseness of implementing a process control scheme seriously and usually relate with products’ quality characteristics. Hence, it is necessary to obtain scale parameters that characterize the variability. In this work, we propose a statistical analysis procedure for batch profiles. First, we divided batch process data into time scaling, location scaling and residuals which characterize the time axis variability, the shift or drift of process variables and the batch profiles with no shift and drift respectively. Next, we analysis the residuals by modified health index and provide analysis and monitor methods for scale parameters. The proposed statistical analysis procedure can not only monitor faults in the process but also investigate the relationship of process variables to achieve the goal of executing process monitoring and process improvement.

並列關鍵字

Batch process ； Fault detection and classification ； Data mining ； Health index ； Hotelling's T-square

參考文獻

[1] Akima, H., (1970). “A New Method of Interpolation and Smooth Curve Fitting Based on Local Procedures,” Journal of the Association for Computing Machinery, 17, 589-602.

[2] Johnson, R. A. and Wichern, D. W. (1982, 2002). Applied Multivariate Statistical Analysis, 5th ed., New Jersey, Prentice Hall.

[3] Kaistha, N., Moore, C. F. and Leitnaker, M.G. (2004). “A Statistical Process Control Framework for the Characterization of Variation in Batch Profiles,” Technometrics, 46, 53-67.

[4] Nomikos, P. and MacGregor, J. F. (1995). “Multivariate SPC Charts for Monitoring Batch Processes,” Technometrics, 37, 41-59.

[5] Nomikos, P. and MacGregor, J. F. (1995). “Multi-way partial least squares in monitoring batch processes,” Chemometrics and Intelligent Laboratory Systems, 30, 97-108.