RAPID FUNCTIONAL-GROUP CHARACTERIZATION OF GAS-CHROMATOGRAPHY FOURIER-TRANSFORM INFRARED-SPECTRA BY A PRINCIPAL COMPONENTS-ANALYSIS BASED EXPERT SYSTEM

A procedure for the on-line characterization of compounds eluting from a gas chromatograph (GC) by automated interpretation of their vapor-phase infrared (IR) spectra is described. Principal components analysis (PCA) is applied to a training set containing two classes of appropriately scaled vapor-phase infrared library spectra. A sigmoid discriminant function is applied to the results of the PCA. The presence of a given functionality in an unknown spectrum is determined by projection of the unknown spectrum onto the factors defined by the PCA and assigning the projected unknown a discrimination score. The product of the discrimination score and the Gram-Schmidt reconstructed chromatogram can be plotted as a chromatographic trace emulating a functional group selective detector. Examples are shown for chemical classes such as alcohols, carboxylic acids, ketones, and aldehydes. A study of the results of baseline correcting the spectra before projection has been completed. Much more accurate results are obtained after baseline correction has been applied.