Integrated machine learning approaches for complementing statistical process control procedures

Abstract

Although statistical process control (SPC) procedures have played a central role in solving quality problems, their effectiveness is yet to be fully realized in the process industry with large volume data, a lot of dimensions of variables, and complex relationships among processes and variables. To complement SPC procedures, we suggest three integrated methods of inductive learning and neural networks for solving the quality problems. First, a feature subset selection method is proposed for reducing variables required for quality control. Second, a clustering inductive learning method is presented for improving the correct prediction rate (CPR) of inductive learning. Third, a pattern detection method is suggested for detection of different patterns comparing reference patterns. Three methods are experimented in two datasets. The results show that the three methods are effective for the multivariate process control.