變號方法的變異點偵測分析

Abstract

不管是在製程監控或其他相關領域，訊號偵測是一個重要的問題。Wang(2012)發展出一套變號方法，將傳統管製圖的統計量與JS 型態統計量相減之後用於偵測失控點。由於此方法不需設立管制界限去偵測訊號，因此不失為一個更有效率的方法用於訊號偵測。我們設計與進行一個模擬研究去比較變號方法與傳統管製圖的表現，而從模擬的結果來看，不管是以MC1或MEWMA統計量為基準，我們發現訊號方法都能更有效率的偵測到失控點。

Signal detection is an important problem in monitoring or other related fields. Wang (2012) proposed a sign change method which based on the difference of the traditional chart statistic and the JS-type chart statistic to detect out-of-control signals. Since this approach is not necessary to construct control limits to detect a signal, it can supply an efficient way for signal detection. We design and conduct a simulation study to compare the performance of the sign change method with those based on the traditional charts in this study. From the simulation result, we show that the sign change method can detect out of control signals more efficiently based on the MC1 or the EMWMA control chart statistics.