Identification of Trend and Shift Variation in Bivariate Process Using Pattern Recognition Scheme

Abstract

In advance manufacturing industry, quality control is an important to meet and achieve customer requirement. Statistical Process Control (SPC) is a tool that used in quality control to monitor and diagnosis the manufacturing process. In real production system there is challenging to monitoring and identifying unnatural process variation (UPV) when it composed two correlated quality variables (bivariate). Meanwhile, most of previous research are focusing on the pattern recognition technique. However, research on the findings of unnatural variation is still not many and it is limited to certain pattern such as sudden shifts pattern. In this study, the pattern recognition for unnatural variation is focusing to identify trend and shifts pattern in bivariate process. The framework of scheme was constructed using Artificial Neural Network (ANN) recognizer and window size will be determined using a different size to achieve 99% recognition accuracy. Thus, performance of the proposed framework able to identify the nine category of control chart pattern that focus on trend and shift pattern. The result demonstrates the proposed approach can recognize control chart pattern effectively. In previous research, it was proven that double stage algorithm able to detect and identify the five basic control chart pattern (Shaban & Shalaby, 2010). The outcome of this study will be helpful for quality practitioner in industry that dealing with a multiple variable in their manufacturing process for realizing accurate monitoring and diagnosis the process.