應用資料包絡法最佳化多品質之實驗設計

Abstract

工業界常使用實驗設計（Design of Experiments, D.O.E.）或田口方法（Taguchi Method）來改善製程以提升產品品質。然而隨著客戶導向時代的來臨，產品的複雜度不斷增加才能符合顧客多樣的需求，因此多品質同時最佳化是目前各公司研發單位非常重視的一個議題。傳統之實驗設計與田口方法僅能最佳化單一產品品質特性，目前有關多品質特性同時最佳化的中外文獻，大部分皆是針對田口實驗提出一套多品質最佳化演算法，罕見有文獻是針對傳統之實驗設計提出多品質同時最佳化演算法。有許多文獻指出田口方法的一些缺失並質疑其有效性，而傳統實驗設計法比田口方法嚴謹且能更有效的最佳化產品之品質特性值。因此，本論文利用傳統之實驗設計法規劃多品質特性實驗並結合資料包絡法(Data Envelopment Analysis, DEA)與反應曲面法（Response Surface Methodology, RSM）提出一套多品質特性最佳化演算法，提供工業界找出最佳之製程參數水準組合。本論文最後以新竹科學園區某半導體公司蝕刻製程的一個案例來驗證本研究所提出的最佳化多品質特性的演算法確實有效可行。

Due to the product design becomes increasingly complicated, more than one quality characteristic must be simultaneously optimized to improve the product quality. Design of experiments（DOE）and Taguchi methods are often applied in industry to develop new product or improve the process. However, both can’t solve the multi-response problem directly. Taguchi methods combined with some mathematical calculations is applied in most literatures about the optimization of multi-response. But several of Taguchi’s data analysis methods are questionable. Classical design of experiments will be applied to obtain the optimal parameter-setting effectively. The purpose of this study is to propose a procedure for finding the optimal parameter-setting of a process or product with multiple responses using DOE combined with Data Envelopment Analysis (DEA) and Response Surface Methodology（RSM）. A case with multiple quality characteristics is carried out in wafer manufacturing in Taiwan to demonstrate the practicability of the proposed procedure.