透過您的圖書館登入
IP:18.188.152.162
  • 學位論文

應用在自相關性資料上的 管制圖無模式設計_以實驗分配為基

The Model-Free Design of Control Charts For Autocorrelated Data Based on the Empirical Distribution

指導教授 : 陳慧芬
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


這篇文章主要研究在具有自我相關品質特性但未知分配以及相關性的製程,我們給定一組資料,已知此組資料服從ARTA (Autoregressive-To-Anything) process,但未知屬於何種特定分配以及相關係數未知,所以利用這組給定的資料建構Empirical distribution,當作此製程的分配。 之前有許多研究建構Empirical distribution(經驗分配),有好幾種不同的方式,包括單純的piecewise-linear建構方式,以及後來延伸到利用指數分配去做為右邊極值 分配的方式,而本篇研究應用相同的方法,利用指數分配作為 分配的方式,建構出雙邊尾巴利用指數分配建構的模式。 這篇文章也將資料著重在於有相關性的部份,利用ARTA process 以及我們建立出來的雙邊經驗分配,建立出有自我相關性的觀察值,接著比較原始製程與我們建立出來分配的觀察值的差異,檢驗利用此方式建構出來的經驗分配,是否能近似於原始資料的分配,以及在不同樣本數建立出來的經驗分配,或是不同強度的自我相關性,產生出來的觀察值與原始分配的差異來做分析,也能夠藉由與原始分配的比較,判斷建立出的經驗分配與原始分配的相似度。

並列摘要


We consider the design of X charts, where the quality characteristic measurements are autocorrelated with and unknown marginal distribution and unknown autocorrelations, but a set of data is given. Specifically, we assume that the data process follows an ARTA (Autoregressive-To-Anything Process). Then we apply the X chart, we need to decide the value of the X-chart design parameters: the sample size n and the control limits. In practice, the data properties are unknown. Here we apply two steps to compute the ARL (Average Run Length). First, we construct an empirical distribution for a given set of data. We use the mixture of the original empirical distribution and exponential distribution (for the tails). Second, we evaluate the performance of the model-free design based on the proposed empirical distribution. The performance measurements are the in-control and out-ofcontrol average run length (ARL). The ARL is the expected number of samples (or observations if taking samples of one) required by the chart to signal.

參考文獻


[3] Cario M. C. and Nelson B. L. (1997). Numerical Methods for Fitting and Simulating
Engineering and Management Sciences Northwestern University, Evanston,
Ransom Variate Generation Using a Piece-linear Cumulative Distribution
[5] Montgomery, D. C. (2000). Introduction to Statistical Quality Control, 5th ed.
John Wiley and Sons, New York.

延伸閱讀