A Computer System For Analyzing And Viewing Complex Plant Data

Researchers at BCD Technologies and the University of Alabama have applied three statistical techniques to the information contained in a large database of measurements from a phosphate ore processing plant. The rust, Principal Components Analysis (PCA), replaces many measured variables with a few fundamental and un-correlated variables. The second technique, Projection to Latent Structures (PLS), is an extension of PCA that explores the relation between variables identified as either independent or dependent. The third technique, Cluster Analysis (CA), is applied to the results of the PCA or the PLS calculations, and identifies individual data clusters that represent uniform operating conditions in the plant. All three statistical methods can be seen as geometrical operations on the data viewed as points in space. A Viewer gives a corresponding graphical presentation of the data points. This paper describes the use of the above three techniques and the Viewer to explore part of the phosphate flotation database.