Comparing Statistical Feature and Artificial Neural Networks for Control Chart Pattern Recognition: A Case Study

Hadiyat, M. Arbi and Prilianti, Kestrilia Rega (2012) Comparing Statistical Feature and Artificial Neural Networks for Control Chart Pattern Recognition: A Case Study. Proceeding : 3rd International Conference on Technology and Operation Management 2012. pp. 83-88. ISSN 978-979-15458-4-6

[img]
Preview
PDF
Hadiyat_Prilianti_Comparing_Abstract_2012.pdf

Download (142Kb) | Preview
[img]
Preview
PDF
Hadiyat_Prilianti_Comparing_2012.pdf - Published Version

Download (334Kb) | Preview
[img]
Preview
PDF
Hadiyat_Prilianti_Comparing_References_2012.pdf

Download (339Kb) | Preview

Abstract

Control chart has been widely used for monitoring production process, especially in evaluating the quality performance of a product. An uncontrolled process is usually known by recognizing its chart pattern, and then performing some actions to overcome the problems. In high speed production process, real-time data is recorded and plotted almost automatically, and the control chart pattern needs to be recognized immediately for detecting any unusual process behavior. Neural networks for automatic control chart recognition have been studied in detecting its pattern. In the field of computer science, the performance of its automatic and fast recognition ability can be a substitution for a conventional method by human. Some researchers even have developed newer algorithm to increase the recognition process of this neural networks control chart. However, artificial approaches have some difficulties in implementation, especially due to its sophisticated programming algorithm. Another competing method, based on statistical feature also has been considered in recognition process. Control chart is related to applied statistical method, so it is not unreasonable if statistical properties are developed for its pattern recognition. Correlation coefficient, one of classic statistical features, can be applied in control chart recognition. It is a simpler approach than the artificial one. In this paper, the comparison between these two methods starts by evaluating the behavior of control chart time series point, and measured for its closeness to some training data that are generated by simulation and followed some unusual control chart pattern. For both methods, the performance is evaluated by comparing their ability in detecting the pattern of generated control chart points. As a sophisticated method, neural networks give better recognition ability. The statistical features method simply calculate the correlation coefficient, even with small differences in recognizing the generated pattern compared to neural networks, but provides easy interpretation to justify the unusual control chart pattern. Both methods are then applied in a case study and performances are then measured.

Item Type: Article
Uncontrolled Keywords: Control chart, pattern recognition, neural network, correlation, back propagation
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: Faculty of Engineering > Department of Industrial Engineering
Depositing User: Kestrilia Rega Prilianti 61150
Date Deposited: 03 Aug 2012 03:04
Last Modified: 03 Aug 2012 03:04
URI: http://repository.ubaya.ac.id/id/eprint/905

Actions (login required)

View Item View Item