應用資料視覺化技術探討PCB新製程品質特性缺陷成因

Improved data visualization techniques for detecting quality problems of PCB new process

林家弘(Jia-Hong Lin)

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

印刷電路板產業可說是一個製程技術發展水準相當高的電子科技產業，在眾多同業相互競爭之下，開發新製程技術提升其產品品質是必然的發展趨勢。目前作法都是工程師憑藉其經驗來選取可能影響品質的控制因子，輔以實驗設計、實驗批驗證，來訂定製程參數值調整範圍，但製程控制因子過多，且彼此間具有相關性，故於選取重要關鍵因子時非常困難。而新製程資料蒐集又有資料量少的原因，於資料分析上會造成傳統統計方法及一般資料探勘方法會造成偏誤。故本研究與蘇星樺(2003)合作，利用資料視覺化方法較不局限於參數維度與筆數多寡的特性，找出製程控制因子間與品質特性的相關性，作為造成品質特性缺陷之重要關鍵因子的選取依據，此成功實務案例，記錄於蘇星樺(2003)中。本研究首先將此視覺化資料探勘過程以系統化架構整理，並提出新的視覺方法以及維度排序方法解決蘇星樺(2003)方法中資料視覺呈現時資料覆蓋與視覺特徵雜亂的問題。另外，本研究提出不同視覺特徵之相關性權重量化指標，強化蘇星樺(2003)對於視覺特徵之相關性上定義的不足之處，幫助工程師能擁有更多有關於控制因子間相關性探討的資訊。

關鍵字

PCB製程品質特性缺陷成因 ； 資料探勘 ； 資料視覺

並列摘要

For companies in a highly-competitive industry such as PCB manufacturing to remain competitive in the market a critical issue is to keep developing new manufacturing processes that have lower cost or higher quality. In an effort cooperated with Su (2003) we have successfully developed a visual data mining process to address an important problem in developing a new manufacturing process: finding the process controlling parameters that affect the quality features of the product. In this research, we outline Su’s visual data mining work using a framework architecture. Then we point out several problems in Su’s work and propose solutions to these problems. First, we discuss the data-overlapping problem in parallel coordinate graph and propose survey plot to overcome this problem. Secondly, we discuss problems with Su’s dimensional layout and propose new method to layout the dimensions in parallel coordinates. Thirdly, we propose different way to obtain a quantitative measure for the visual patterns shown in a parallel coordinates. We use the same data in Su’s work and compare the results to show the improvement.

並列關鍵字

visual data mining ； data visualization ； PCB manufacturing quality problems

參考文獻

DAN, B. and ARMIN, S., (2002), “Data Mining for Improving a Cleaning Process in the Semiconductor Industry”, IEEE Transactions On Semiconductor Manufacturing, VOL. 15, NO. 1, February, pp. 91-101.

FAYYAD, U. M., (1997), ”Data Mining and Knowledge Discovery: Making Sense out data”, IEEE Expert, Vol. 11, No.5, October, pp. 20-25.

GIRARDIN, L. and BRODBECK, D., (2001) “Interactive Visualization of Prices and Earnings around the Globe”, IEEE Computer Society Technical Committee on Visualization and Graphics, October 22-23, San Diego, California, USA, pp. 1-10.

GRUPE, F. H. and OWRANG, M. M., (1995), “Data Base Mining Discovering New Knowledge and cooperative Advantage”, Information Systems Management, Vol. 12, No. 4, pp.26-31.

PHILIP, B., (2002) “Picture this-Drowning in data? New visualization techniques could help. Philip Ball discovers, among other things, how to plot a seven-dimensional graph”, Nature, Vol. 417, July, pp. 11-13.