多變量自我相關性資料的製程監控

Abstract

近幾年,統計製程管制的另一個重要焦點在於發展監控具有自我相關性的資料的控制圖。 本篇文章中, 我們研究多變量自我相關性資料的控制圖。 將介紹多變量特殊原因管制圖及推導其連串長度分配。 我們並且針對VAR(1) 的製程, 將多變量特殊原因管制圖與Kramer和Schmid (1997) 所提的時間數列資料的MEWMA 控制圖和將殘差值應用在Lowry 等人所提的多變量EWMA 管制圖做ARL值的比較。 最後將方法應用在兩個實際例子上 。

In recent years, an important focus of research in statistical process control (SPC) is control charts for autocorrelated data. In this paper, a multivariate special-cause chart (MSCC) is introduced. The run length distributions of the multivariate special-cause chart are derived for ARMA process. We also compare this chart on the basis of the average run length (ARL) with the MEWMA control chart for time series (Kramer and Schmid, 1997) and the MEWMA control chart applied to residuals for VAR (1) processes. Finally, MSCC is demonstrated with two real examples.