權重設定下之多反應變數最佳化演算法

Abstract

隨著時代快速演進，全球化的趨勢使得產業競爭越來越激烈，不論是產品的設計或製造的過程，已無法使用單一品質特性來衡量產品的優劣，而必須使用多個品質特性才能衡量出產品的品質。目前雖有文獻提出許多利用田口方法(Taguchi Method)或實驗設計(Design of Experiments, D.O.E)來同時最佳化多個品質特性或多個反應變數的方法，但這些方法大多是先將多個反應變數整合成權重相同的單一指標(Multiple Performance Characteristics Index, MPCI)，再最佳化此指標，然而，在探討多反應變數最佳化問題時，各反應變數的重要程度並不相同，且整合成單一指標後，並未檢視各個反應變數的表現是否與其重要性一致。本論文利用模糊理論(Fuzzy Theory)、望想函數(desirability function)及自組性演算法(Group Method of Data Handling, GMDH)結合實驗設計手法，發展出一套多反應變數同時最佳化演算法，以找出一組能表現各反應變數重要性之最佳因子水準組合。本論文最後以新竹某廠商的散熱系統之案例，來說明本論文提出之方法確實有效及可行。

With the rapid evolution of the times, the trend of globalization makes more competition in every industry, the design of products also become more and more complicated, consequently, one single quality characteristic is not enough to mea- sure the total quality of a product and the optimization of multi-response becomes increasingly important. Many studies developed methods to design the experime- nts to simultaneously optimize multiple quality characteristics using Taguchi met- od or Design of Experiments (D.O.E). Those studies usually integrated multiple response variables into one index and then optimizing the composite index, however, the importance of each response variable may be different, composite integrating these responses into one index without considering the relationship between the importance and the performance of each response composite may not be appropriate. Hence, the main objective of this study is to develop a mult-respo- nse optimization algorithms using Fuzzy Theory, desirability function and Group Method of Data (GMDH) combined with DOE to determine the optimal settings for the factor-level. A case study of the cooling system is used to demonstrate the effectiveness and feasibility of the proposed procedure.