An Approach for Control Chart Pattern Recognition using the Fuzzy ARTMAP Artificial Neural Networks with Improved Efficiency

Published in: Megaprojects: Building Infrastructure by Fostering Engineering Collaboration, Efficient and Effective Integration and Innovative Planning: Proceedings of the 10th Latin American and Caribbean Conference for Engineering and Technology
Date of Conference: July 23-25, 2012
Location of Conference: Panama City, Panama
Authors: José Antonio Vázquez-López
Susana Goytia-Acevedo
Ismael López-Juárez
Armando J. Ríos-Lira
Refereed Paper: #145

Abstract

The use of Control Charts (CC) in manufacturing processes is a common technique to monitor the quality of the production. The production variables are monitored to preserve the process under statistical control and also to detect any special variation. Causes for special variation are diverse such as change in material processing, changing the machine operator, changing the machine itself, etc. These changes are typically showed in CC and interpreted by trained personnel to take appropriate actions to get the process under statistical control. In this paper we introduce an approach to recognise and analyse statistical patterns using Artificial Neural Networks (ANN’s). The approach is based on the FuzzyARTMAP (FAM) network whose parameters are selected on-line depending on the encountered probability distribution and whether special or non-special patterns are encountered; hence, the mechanism for selection is driven by the type of probability distribution. In terms of the network parameter selection, their value is not predefined but established a priori based on a sensitivity analysis for improved the efficiency. Experimental results showed that the range selection of these parameters is very important to improve the efficiency of the FAM network and to establish a robust method to effectively recognise CC patterns in Statistical Process Control (SPC).