THE USE OF PRINCIPAL COMPONENT ANALYSIS AS A DATABASE MINING TOOL FORTHE EXPLORATORY DIAGNOSIS OF CHROMATOGRAPHIC PROCESSES

The work reported in this paper examines the use of principal componen t analysis (PCA), a technique of multivariate statistics to facilitate the extraction of meaningful diagnostic information from a data set o f chromatographic traces. Two data sets mimicking archived production records were analysed using PCA. In the first a full-factorial experim ental design approach was used to generate the data. In the second, th e chromatograms were generated by adjusting just one of the process va riables at a time. Data base mining was achieved through the generatio n of both gross and disjoint principal component (PC) models. PCA prov ided easily interpretable 2-dimensional diagnostic plots revealing clu sters of chromatograms obtained under similar operating conditions. PC A methods can be used to detect and diagnose changes in process condit ions, however results show that a PCA model may require recalibration if an equipment change is made. We conclude that PCA methods may be us eful for the diagnosis of subtle deviations from process specification not readily distinguishable to the operator.