應用灰關聯分析與等級離散統計量於等級類別資料之製程參數最佳化

Abstract

隨著產品或製程複雜度的增加，在進行產品品質改善的實驗時，品質特性常須以等級類別資料的型式呈現，而過去所發展之等級類別資料製程參數最佳化之方法，大多牽涉到複雜的數學模式或無法完全正確地衡量資料中的變異與接近目標值的程度。所以本研究分別利用灰關聯分析與等級離散統計量來衡量資料中的位移效應與離散效應，再結合反應曲面法與望想函數構建出一套等級類別資料之製程參數最佳化方法，以達到使平均值接近目標值及縮小變異的目的。最後以一個半導體離子植入製程實例來說明如何應用本研究所發展的演算流程及驗證其有效性。

Due to the increasing complexity of the product design, the quality response is often measured by ordered categories when using design of experiments for quality improvement. Despite their extensive use in optimizing process parameters for a qualitative quality response, conventional methods are unnecessarily complex or inaccurate. Therefore, this study develops a procedure capable of optimizing the location and dispersion effects simultaneously via setting the process parameters in a continuous level for an ordered categorical response. This novel method applies grey relational analysis and ordinal dispersion statistics to estimate the location and dispersion effects, respectively. Response surface method and desirability function are then performed to obtain the optimal factor/level combinations. Finally, an ion implantation process owing to a Taiwan’s IC fabrication manufacturer is utilized to demonstrate the feasibility and verify the effectiveness of the proposed procedure.