資料視覺化探勘輔助偵測PCB新製程之不良成因

Data Visualization assist in detecting causing factor of low-yield for PCB new process

蘇星樺(Hsing-Hua Su)

指導教授 ： 孫天龍

元智大學/工程學院/工業工程與管理學系/碩士(2003年)

摘要

印刷電路板產業(Printed Circuit Board, PCB)是一個發展技術成熟的電子產業，因此在面臨眾多的同業廠商激烈競爭之情況下，開發新製程技術與提升產出之品質是一個必要的策略趨勢。目前工程師往往根據經驗選擇可能影響品質的控制因子，利用實驗設計與實驗批來驗證。但因製程中的控制因子過多，且因子又會相互影響，故要從中挑選重要因子實非簡單之事。且因新製程中蒐集的資料有資料量少、變數多的特性，導致一般以電腦演算法(computer algorithm)為中心的資料探勘技術作分析非常困難。而資料視覺方法有不需對母體做假設或定義，以及在探勘過程可結合工程師的專業意見等優點。故在本研究以資料視覺方法 — 平行座標圖(Parallel Coordinates)做為找出影響製程品質特性重要因子的工具，並討論在此視覺方法呈現下的視覺特徵與其代表的製程意義。此外，利用維度配置的手法，明確的顯示圖形中的視覺特徵，以及透過本研究中的分群方法與視覺呈現設計觀察因子間的關係，以做為工程師在製程研發與改善上的決策參考依據。
關鍵字：視覺化資料探勘、平行座標圖、PCB製造品質問題

關鍵字

PCB新製程技術 ； 平行座標圖 ； 資料視覺

並列摘要

Developing a new manufacturing process with lower cost and higher quality is critical for PCB manufacturing companies to remain competitive in the market. One challenging task in developing a new manufacturing process is to find out those process controlling parameters that affect the quality features of the product. This task is difficult due to the large number of controlling parameters in the process and the complicated interactions between these parameters. Recently, many researchers have proposed to use data mining techniques to analyze data collected from the manufacturing process to find out the causing factors of a fault problem. For a newly developed process, however, since the data must be collected from experimental batches and the company has to pay extra money to run the experimental batches, the number of data collected is usually less. Consequently, the data mining algorithms will have problems to apply. Visualization, on the other hand, is more suitable in this case. This research presents a visual data mining approach for analyzing the small amount of data collected from a newly designed PCB manufacturing process. We first discuss the input data representation, the visualization design, the visual patterns shown in the visual images, and the manufacturing meanings of the visual patterns. Then we present the visual data mining process that is used to find out the controlling parameters that affect the quality features of the product.
Keywords: visual data mining, parallel coordinates, PCB manufacturing quality problems

並列關鍵字

PCB new process technique ； Parallel Coordinates ； Data Visualization

參考文獻

BRAHA D. and SHMILOVICI A. (2002), “Data Mining for Improving a Cleaning Process in the Semiconductor Industry”, IEEE Transactions on Semiconductor Manufacturing, Vol. 15, No. 1, February.

FAYYED U., PIATETSKY-SHAPIRO G.., and SMYTH, P. (1996), “The KDD Process for Extracting Useful Knowledge from Volumes of Data.”, Communication of ACM, Vol.39, No.11, pp.27-34.

FAYYED U., GRINSTEIN G. G. & WIERSE A. (2002), Information Visualization in Data Mining and Knowledge Discovery.

GARDNER M. and BIEKER J. (2000), “Solving Tough Semiconductor Manufacturing Problems Using Data Mining”, IEEE/SEMI Advanced Semiconductor Manufacturing Conference.

INSELBERG A. (1985), “The Plane with Parallel Coordinates,” The Visual Computer 1, pp.69-91.