On integration of statistical process control and engineering process control: a neural network application

Abstract

In recent years the importance of quality has become increasingly apparent. Quality control in manufacturing has moved from detecting non-conforming products through inspection to continuously reducing variability in product performance and production process. Two existing fields which have been contributing to quality are statistical process control (SPC) and engineering process control (EPC). In the past, these fields have been developed in relative isolation from one another, resulting in process control engineers trained primarily in classical control theory and almost no training in statistics, data analysis and process noise control. On the other hand, statisticians have a poor understanding of process dynamics and classical control theory, but they have excellent training in the analysis of discrete data, the design of experiments and the methods for empirical process modeling (Gupta and Kumar, 1991). Both methodologies aim to bring all the process levels to their targets with minimum variability. Statistical process control commonly assumes that under normal conditions the process variability is driven by common causes, the effect of which is impossible or too expensive to reduce, and that an increase in variability or unfavorable changes in the process mean are due to special causes.