Design Methodology of Modular-Ann Pattern Recognizer for Bivariate Quality Control

Abstract

In quality control, monitoring unnatural variation (UV) in manufacturing process has become more challenging when dealing with two correlated variables (bivariate). The traditional multivariate statistical process control (MSPC) charts are only effective for triggering UV but unable to provide information towards diagnosis. In recent years, a branch of research has been focused on control chart pattern recognition (CCPR) technique. However, findings on the source of UV are still limited to sudden shifts patterns. In this study, a methodology to develop a CCPR scheme was proposed to identify various sources of UV based on shifts, trends, and cyclic patterns. The success factor for the scheme was outlined as a guideline for realizing accurate monitoring-diagnosis in bivariate quality control.