Principal component analysis (PCA) has been used for the modeling of nonlinear chromatography under overload conditions. A 10-fold range of crude erythromycin samples were loaded onto columns with different stationary-phase chemistries (2 polystyrene, 1 methacrylate) in direct proportion to the bed volumes. The elution profiles indicated slightly concave isotherms for the polystyrene supports and a convex Langmuirian isotherm for the methacrylic support used. The principal component models accounted for over 98% of the original variance in the data for all three columns and were able to give excellent models of complete chromatograms in the absence of first-principle models or physicochemical data. Correlations between sample mass and the principal component scores were made for each that were consistent for the column types despite the different geometries and stationary phases. Linear relationships with high correlation coefficients were observed when the scores of the same principal component were compared between columns. Such correlations offer considerable potential for modeling of nonlinear chromatography.

The work reported in this paper examines the use of principal component analysis (PCA), a technique of multivariate statistics to facilitate the extraction of meaningful diagnostic information from a data set of chromatographic traces. Two data sets mimicking archived production records were analysed using PCA. In the ®rst a full-factorial experimental design approach was used to generate the data. In the second, the chromatograms were generated by adjusting just one of the process variables at a time. Data base mining was achieved through the generation of both gross and disjoint principal component (PC) models. PCA provided easily interpretable 2-dimensional diagnostic plots revealing clusters of chromatograms obtained under similar operating conditions. PCA methods can be used to detect and diagnose changes in process conditions, however results show that a PCA model may require recalibration if an equipment change is made. We conclude that PCA methods may be useful for the diagnosis of subtle deviations from process speci®ca-tion not readily distinguishable to the operator.

No abstract