TABLE OF CONTENTS In order to meet quality, safety, efficiency, cost, and environmental targets in the process industries, it is often necessary to combine process control and monitoring methods [8-1]. Process control techniques were discussed in earlier chapters of this book, so this chapter focuses on monitoring techniques. Traditionally, statistical process control (SPC) charts such as Shewhart, CUSUM, and EWMA charts [8-2] [8-3] have been used to monitor industrial processes for the purpose of improving product quality. Such techniques are well developed for manufacturing processes and widely used in univariate systems in which a single process variable is monitored. However, industrial process performance usually includes more than one process variable, and univariate SPC charts do not perform as well for these multivariate systems [8-3] [8-4], [8-5]. For example, in chemical and biological systems, hundreds of variables can be recorded regularly in a single operating unit, resulting in a large data set that is hard to analyze with univariate methods. Furthermore, because many of the variables may be correlated, they must be considered all together rather than individually. To meet these requirements, multivariate statistical process control (MSPC) methods can be employed to reduce the dimension of the large raw data set while extracting useful information such as the existence of faults or abnormalities.
Since the late 1980s, industry has...
