田口方法望大望小品質特性SN比之研究

Abstract

田口品質工程的基本理念在於強調品質是設計出來的，而非檢驗出來，應用參數設計的方法，使在高效率及低成本的情況下，迅速地改善品質。田口方法在品管理念上的貢獻毋庸置疑，但在方法上卻有一些缺失。例如它在處理望大、望小品質特性時，若根據其定義的SN比來選取水準，會發生一些問題： 一、望大、望小之SN比不能反應因子對σ2的顯著性，可能導致選錯最佳因子水準。 二、SN比沒有對u與σ2協調 (trade off) 的作用，它偏向以μ來選取水準，而忽視同樣重要的σ2。 本論文旨在研究田口望大、望小SN比的缺失，並利用多屬性決策方案之歐幾里得距離近似衡量法，用另一個角度重新導出因子水準選取的函數指標。最後，並以IC在金屬化玻璃板之黏合力實驗來驗證新的指標的可行性，以及說明新的指標因改善上述田口望大、望小SN比的缺點，而返出了較為合理的因子水準。

The basic concept of Taguchi method is that we can iimprove the qualities of products efficiently with low cost by means of parameter design. Although Taguchi method has many advantages in practical application, there are still some drawbacks. For example: 1. The signal to noise ratio (SN) for smaller-is-better (SNs) and larger-is-better (SNL) quality properties only emphasize on u. They can not really reflect the importance of σ2. 2. In the case where u and σ2 conflict, SNs and SNL lose the ability of trading off between u and σ2. The values of these SN ratios are affected mainly by u and neglect σ2, which is as important as u. The objective of this thesis is to study the drawbacks of SNs and SNL. Euclidean distance proximity measurement of multiattribute decision making is applied to develop two new indexes to replace SNs and SNL. An example of maximizing bond strength when mounting an IC on a metallized glass substrate is illustrated to show the feasibility of the new indexes developed by this study.