統計管制圖是監控製程中產品品質穩定性的重要工具。隨著現代製程的進步，製程中經常需要監控多個的品質特性，因此多變量統計製程的設計日益重要。為能設計一種穩健且能監測多變量平均向量微量改變的製程管制方法，本文建議混合使用多變量指數加權移動平均(MEWMA)及累和管制圖(MCUSUM)，其中各品質特性之邊際分布為常態分布，但是應用關連結構模型描述品質特性之聯合分布。本文藉由蒙地卡羅模擬比較所提方法與MEWMA及MCUSUM管制圖在品質特性失控時所需的平均連串長度。最後利用一個在計算機製程中的一組真實數據展示所提方法的實際應用。;Statistical control charts are important tools for monitoring the stability of product quality in manufacturing processes. With advancements in modern processes, it is of increasing importance to monitor more than one quality characteristic simultaneously, hence emphasizing the need for designing multivariate statistical process control methods. To develop a robust method for detecting subtle changes of mean vectors in multivariate processes, this paper proposes a hybrid approach using Multivariate Exponentially Weighted Moving Average (MEWMA) and Multivariate Cumulative Sum (MCUSUM) charts. The marginal distributions of quality characteristics are assumed to be normal distributions, while a structural model is employed to describe the interrelations among quality characteristics. Through Monte Carlo simulations, the performance of the proposed method is compared with that of MEWMA and MCUSUM control charts on the average run length required when multivariate quality characteristics are out-of-control. Finally, the practical application of the proposed method is illustrated by a real-world dataset arising from a computer manufacturing process.
