利用灰關聯分析辨認多變量製程管制失控之特殊變因

Abstract

在現今繁雜的製造環境中，一個產品往往同時包含了數個品質特性，而這些品質特性間又彼此相關，這使得多變量管制方法越趨重要。自1947年Hotelling利用c2分配構建了第一個多變量管制圖後，陸續有許多關於多變量統計製程管制方面的研究發表。其中包括了Hotelling T2管制圖的修正及簡化、確認失控的品質特性因子及管制界限的制定等，多變量管制圖的理論架構可說是日趨完備。然而，因為多變量管制圖所根據之數學理論太過深奧或是推導程序太過繁雜，使得在辨認導致多變量製程失控的特殊變因上，仍有可改進之處。有鑑於此，本研究利用灰色理論(gray theory)中之灰關聯分析作為辨認導致多變量製程失控之特殊變因的工具，提出一個較簡易的多變量製程診斷程序，以有效地辨認出導致製程失控之特殊變因。本研究最後並以實例說明如何使用本研究所發展之多變量製程失控特殊變因的辨認程序。

Stringent market competition has driven manufacturers to enhance product quality. A product's quality is usually evaluated by several quality characteristics and these characteristics are often correlated. In this case, multivariate process quality control methods are useful. After Hotelling developed the first multivariate control chart, T2 control chart, based on c2 distribution in 1947, there have been a lot of related papers published, including the modification and simplification of Hotelling's T2 control chart, identification of the out-of-control quality characteristics, setting of control limits, and etc. However, because most of the multivariate control charts involve difficult mathematical theories or complicated computation, the identification of special causes accounting for the out of control situation in the multivariate process control becomes very difficult. The purpose of this study is to employ Gray Relation Analysis of Gray Theory to efficiently identify the special causes accounting for the out of control situation in the multivariate process control. Finally an example is illustrated to demonstrate the proposed procedure.