The performance evaluation of the bivariate EWMAcontrol chart using CARL distribution and EPC

Selly Acita, Muhammad Mashuri*, Dedy Dwi Prastyo

*Corresponding author for this work

Research output: Contribution to journalConference articlepeer-review

Abstract

In general, control charts are developed with the assumption that the critical quality of a production process is normally distributed with known parameters. However, in practice, critical quality is not always normally distributedand process parameters are typically unknown. In such a case, it is necessary to estimate the parameters. One of the measuresto evaluate the performance control charts is Average Run Length (ARL). When the process parameters are estimated, the run-length follows its conditional distribution so-called Conditional Average Run Length (CARL). In this study, the performance of the Bivariate Exponentially Weighted Moving Average(BEWMA)control chart will be evaluated by considering the practitioner to practitioner variability using the CARL distribution and the Exceedance Probability Criterion (EPC). The value of CARL is calculated using the Markov Chain method. The EPC is used to evaluate practitioner to practitioner variability that is closely related to parameter estimation.The results show that to guarantee the in-control performance Phase II chart based on EPC, the large size of observations in Phase I data is needed. However, in practice, it is difficult to collect such a huge size ofdata in Phase I. Therefore, to produce the best in-control performance with available Phase I, the control limits are adjusted.

Original languageEnglish
Article number012053
JournalJournal of Physics: Conference Series
Volume1538
Issue number1
DOIs
Publication statusPublished - 19 Jun 2020
Event3rd International Conference on Combinatorics, Graph Theory, and Network Topology, ICCGANT 2019 - East Java, Indonesia
Duration: 26 Oct 201927 Oct 2019

Fingerprint

Dive into the research topics of 'The performance evaluation of the bivariate EWMAcontrol chart using CARL distribution and EPC'. Together they form a unique fingerprint.

Cite this