Biplot analysis of process systems
Plant data are collected in ever-increasing volumes in modern process industries and play an essential role in monitoring of product quality, control and continuous process improvement. In this paper a modern methodology of biplots is introduced as a means for analysing the behaviour of process systems by enhancing the aforementioned multivariate techniques. This sophisticated biplot methodology is highly graphical in nature and allows for the projection of high-dimensional data to a low-dimensional subspace leading to graphical displays for visualisation by a human operator. Biplots can be regarded as multivariate extensions of ordinary scatterplots where multidimensional sample data points are displayed simultaneously with information of all the variables concerned. In addition, two case studies demonstrate how class separation and multidimensional overlap can be quantitatively expressed by superimposing alpha-bags on biplots.