摘 要 產品或製程之多重品質特性最佳化的決定，對提昇產品品質或製程能力是相當重要的議 題。針對這項議題，本研究先利用田口方法中處理單品質特性的方法，分別計算出各品質特性 的S/N 比，再運用灰關聯分析將S/N 比轉為灰關聯係數，最後結合模糊推論系統導出代表多 重品質特性之衡量指標 (MPCI)，利用其數值愈高愈好的特性，找出最佳參數組合，再由多變 量分析找出同時影響兩種品質特性的顯著因子，以進行包含兩種以上品質特性的最佳參數組合 之模擬驗證，俾驗證最佳參數組合之正確性。本研究以液晶顯示器製程中TFT 的金屬鍍膜製 程為案例，來闡述所提最佳化程序與模式之數值分析，驗證本研究之有效性與實用性，幫助 個案公司節省實驗成本、縮短新產品由實驗階段導入生產階段之時程。

Abstract The traditional Taguchi method mainly aims at the optimization of a single quality characteristic. In fact, there are usually two or more quality characteristics contained in a product or a manufacturing process. In general, the engineer’s experiences are crucial to the determination of factor levels while dealing the issue of multiple quality characteristics. The uncertainty or ambiguity, however, happens frequently for the various standards from different engineers. In addition, when there are more quality characteristics considered or the quality characteristics are more relevant, it would be more difficult for the engineer either to decide the best combination of factor levels or to solve the problem of multiple quality characteristic conflicts. Consequently, determining the product optimization of multiple quality characteristics has become a very important issue in improving the enterprise product quality. The optimization of multiple quality characteristics in the product or manufacturing process is critical for promoting the product quality or the process capability. Therefore, the research first applied Taguchi method considering only one quality characteristic to separately calculate the S/N rations which were then transferred to the Grey Relational Coefficient by the Grey Relation Analysis. Next, the Fuzzy Inference System was utilized to find the Multiple Performance Characteristics Index (MPCI). The higher the values were, the better the parameter combination would be. After that, the main factors that simultaneously affected two quality characteristics were found by executing Analysis of Variance, which verified the best parameter combination for two or more quality characteristics. Finally, taking the TFT metal coating process in the LCD manufacture as an example, the present research explained the numerical analysis of the optimization procedure and pattern to demonstrate its effectiveness and practicality. It is hoped that this study will help the company save the cost for experiment and time for moving the stage from experiment to production.