應用主成份分析方法於多重品質製程最佳化之研究

Abstract

以往田口方法應用於線外品管時，大部分是處理單一品質特性製程的 最佳化，對於在同一製程中有多個品質特性同時最佳化的問題則很少提及 。因此，一般在考慮多品質特性最佳化的問題時，若遇到各單品質最佳因 子水準衝突的情形，大多停留在以工程判斷的階段。目前所發表之文獻雖 提出了一些多重品質特性同時最佳化的方法，但這些方法不是權重值不易 決定，就是徒增問題之複雜度，工程人員並未能有效且容易地解決問題。 本研究提出以主成份分析方法來解決多重品質最佳化的問題。此法是先將 個別品質特性損失標準化，以消除損失函數因不同的衡量單位所造成之差 異，使不同品質特性之損失能處在同一水準上作比較，再運用主成份分析 方法依多個品質特性間的相關性來合併品質特性，最後將這些合併後的新 綜合品質特性應用田口線外製程最佳化的精神，找出最佳的因子水準組合 ，以達到多重品質特性最佳化的目的。本研究並以兩個實例來驗證所提演 算法之可行性與優越性。 Most applications of Taguchi methods only deal with the optimization of asingle-response process or product. Handling the more demanding multi-responsecases is hardly ever considered in the literature. Therefore, when there are more than one response in a process,the usual recommendation is "to be left toengineering judgement". Hence the combination levels of factors chosen in thiscase is often too subjective. At present, few papers have developed some techniques for simultaneous multi-response optimization, however, most of these techniques are either trouble some to determine the proper weight foreach response or too complicated for engineers to use. In this thesis, an effective procedure is proposed to optimize the multi-response process on the basis of the principal components. With the principal component analysis,a set of the original responses can be transformedinto a set of uncorrelated components. The optimal settings of the design parameters for multi-response problems can then be obtianed easily by Taguchi' sidea. The effectiveness of the proposed procedure is evaulated by two case studies.