Please use this identifier to cite or link to this item:
http://kmutnb-ir.kmutnb.ac.th/jspui/handle/123456789/581| Title: | Analysis and Numerical Average Run Length of MEWMA, EEWMA, and NEEWMA Control Charts using the Markov Chain Approach การวิเคราะห์สูตรและคำนวนเชิงตัวเลขค่าความยาวรันเฉลี่ยของแผนภูมิควบคุม MEWMA EEWMA และ NEEWMA โดยวิธีลูกโซ่มาร์คอฟ |
| Authors: | APITAD KRAICHOK อภิธัช ไกรโชค SAOWANIT SUKPARUNGSEE เสาวณิต สุขภารังษี King Mongkut's University of Technology North Bangkok มหาวิทยาลัยเทคโนโลยีพระจอมเกล้าพระนครเหนือ YUPAPORN AREEPONG ยุพาภรณ์ อารีพงษ์ swns@kmutnb.ac.th yupaporna@kmutnb.ac.th |
| Keywords: | average run length Markov chain approach exponentially weighted moving average control chart statistical process control |
| Issue Date: | 17-Nov-2026 |
| Publisher: | King Mongkut's University of Technology North Bangkok |
| Abstract: | This research presents a method for computing the average run length (ARL) of the Modified Exponentially Weighted Moving Average (MEWMA), Extended Exponentially Weighted Moving Average (EEWMA), and New Extended Exponentially Weighted Moving Average (NEEWMA) control charts, which extend the Exponentially Weighted Moving Average (EWMA) control chart and are efficient in detecting small shifts in a process. The ARL is computed by the Markov Chain Approach (MCA), which relies on the cumulative distribution function and on the transitions between in-control and out-of-control states. The monitoring statistics of these charts are formed from more than one observation, so that the transition probability is governed by a joint rather than a marginal distribution and the classical construction cannot be applied directly. Two approaches are therefore proposed: the Joint Distribution Approach (JDA), which evaluates the joint distribution exactly and is developed for the MEWMA chart, and the Point Estimator Approach (PEA), which replaces the lagged observations by an estimator of the process mean and extends to a statistic of any lag depth. Closed-form expressions for the ARL and the standard deviation of the run length are derived, and the process is assumed to follow continuous symmetric, continuous asymmetric, and discrete distributions. Accuracy is assessed by five measures of error together with the computation time required. The proposed method agrees closely with Monte Carlo simulation, the two never differing by as much as one run-length unit under a standard normal process, and its computation time is fixed by the number of partitions of the state space alone, whereas the time required by simulation rises with the length of the runs generated. The method is further used to explain chart behaviour under each form of process distribution, where the effect of the second and third weighted parameters is found to reverse between distributional families, and the charts are applied to four real-world datasets. |
| URI: | http://kmutnb-ir.kmutnb.ac.th/jspui/handle/123456789/581 |
| Appears in Collections: | FACULTY OF APPLIED SCIENCE |
Files in This Item:
| File | Size | Format | |
|---|---|---|---|
| s6704051912011.pdf | 3.57 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.