PENGEMBANGAN DIAGRAM KONTROL MEWMA BERBASIS MODEL UNTUK PENGAMATAN TAK RANDOM

WOROROMI, JONATHAN K. (2016) PENGEMBANGAN DIAGRAM KONTROL MEWMA BERBASIS MODEL UNTUK PENGAMATAN TAK RANDOM. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
1309301005-Abstract.pdf - Published Version

Download (392kB) | Preview
[img]
Preview
Text
1309301005-Phd Thesis.pdf - Published Version

Download (4MB) | Preview
[img]
Preview
Text
1309301005-conclusion.pdf - Published Version

Download (382kB) | Preview

Abstract

Proses produksi pada umumnya melibatkan lebih dari satu karakteristik kualitas (yang saling berkorelasi atau berautokorelasi). Karakteristik ini dikenali sebagai pola sistematik atau pola data tak random. Pola data tak random dalam proses multivariat tidak hanya berpengaruh pada penentuan sensitivitas batas kontrol yang bersesuaian dengan jenis diagram kontrol multivariat, tetapi juga pada kesalahan pendeteksian outlier atau assignable cause lainnya dalam proses multivariat. Diagram kontrol Multivariate Exponential Weighted Moving Average (MEWMA) konvensional pada umumnya mampu bekerja pada shift-shift kecil sampai moderat, dan tidak optimal bila bekerja pada suatu data yang bersifat tak random dengan pola sistematik. Sehingga diperlukan pengembangan suatu diagram kontrol MEWMA yang juga mampu bekerja optimal pada pola data tak random. Penelitian dalam disertasi ini bertujuan untuk mengembangkan suatu diagram kontrol MEWMA baru mampu meningkatkan sensitivitas diagram kontrol MEWMA konvensional terhadap efek pola data tak random. Diagram kontrol MEWMA yang dikembangkan ini merupakan diagram kontrol MEWMA baru berbasis residual dari model Vektor Auto-Regresif (VAR) yang mengadopsi Genetic Algorithm (GA). GA tidak hanya memperbaiki kinerja diagram kontrol MEWMA dalam mendeteksi outlier pada kenaikan-kenaikan shift kecil dan moderat, tetapi juga bermanfaat dalam mendeteksi outlier pada kenaikan-kenaikan shift besar dalam proses vektor mean. Hasil evaluasi terhadap kinerja diagram kontrol MEWMA berbasis model melalui GA dan penaksir Least Squared (LS) menunjukkan perbaikkan kinerja diagram kontrol dalam penanganan efek non normalitas yang berasal dari observasi tak random. Secara analitik, ditunjukkan bahwa efek shift dalam vektor mean dan variabilitas yang terjadi dalam suatu proses dapat dikontrol dan dievaluasi melalui matriks koefisien. Secara numerik, matriks koefisien ===================================================================================================== Generally, a process of production involves more than one quality of characteristics (correlated or autocorrelated). These qualities of characteristics are known as systematic patterns or non random data patterns. Impact of a non random data patterns in the multivariable process is not affected by sensitivity of an associate control limits determination only, but also by fault detection of outlier or from other assignable causes in multivariate process. A conventional multivariate exponential weighed moving average (MEWMA) works on small shifts to moderate shifts level, but it does not optimally work on a non randomized data patterns with a systematic patterns. Therefore, it has been required to develop a new MEWMA control charts that can work optimally on non random data. The aim of the research in this dissertation is to develop a new MEWMA control charts in order to modify the conventional MEWMA due to its sensitivities to effect of the non random data patterns. This new developed MEWMA control chart is MEWMA that based on residuals from a Vector Auto- Regressive (VAR) model with Genetic Algorithm (GA) as an unbiased estimator. The GA estimator was used to improve not only the MEWMA control charts for outlier detection on small to moderate shift increaments, but also large shift increaments in mean vector processes. Due to the result of MEWMA control charts based model evaluations by using of GA and Least Squared (LS) estimator, the performances of the control charts were improved according to non-normality effects from non random observation. Analytically, the occurance of the shift effects on mean vector and variability in a process were controllable and could be evaluated by a coefficient matrices. As the main parameter of VAR(p) models,

Item Type: Thesis (Doctoral)
Additional Information: RDSt 519.86 Wor p
Uncontrolled Keywords: Diagram kontrol MEWMA, Vektor Auto-Regresif (VAR), Pola data tak random, dan Genetic Algorithm (GA)
Subjects: Q Science > QA Mathematics
Q Science > QA Mathematics > QA75 Electronic computers. Computer science. EDP
Divisions: Faculty of Mathematics and Science > Statistics > (S3) PhD Theses
Depositing User: Mr. Tondo Indra Nyata
Date Deposited: 03 Jan 2017 02:28
Last Modified: 27 Dec 2018 08:37
URI: http://repository.its.ac.id/id/eprint/1244

Actions (login required)

View Item View Item