Special-purpose ART-based identification system for statistical process control

Abstract

Although general-purpose pattern identification systems (PIS) have been shown to be useful in identifying a variety of nonrandom patterns, it was achieved at the price of losing the ability to identify certain details in individual pattern classes. In this paper, a special-purpose ART-based PIS for quality control charts used in statistical process control (SPC) is presented. The special-purpose system is proposed to compensate for the limitation of a general-purpose system proposed previously. The adaptive resonance theory (ART) is adopted because it enables fast and cumulative learning which is crucial for real-time, on-line applications. A prototype special-purpose PIS was developed for cyclical data which occur frequently in manufacturing environment. The performance was measured by type I and type II errors as well as by average run length. Simulation shows that the special-purpose ART-based PIS has significant improvement over the general-purpose ART-based PIS.