Automated Chromatographic Data Interpretation Using an Expert System

Abstract

An expert system is developed for automated analysis of gas chromatograms of commercial polychlorinated biphenyl (PCB) mixtures (Aroclors) and Aroclor mixtures. Standard chromatographic data analysis methods are incorporated into the expert system. Principal component regression (PCR) pattern recognition methods are also developed for chromatographic analysis and incorporated into the expert system. The expert system strategy uses knowledge about both of these data processing methods and the knowledge generated about the sample to create a final determination from the results of the two data processing methods. Increases in accuracy and reproducibility are realized through integration of several data processing techniques. This work demonstrates the need for pattern recognition data processing techniques to replace the intuitive pattern recognition performed by chemists on multicomponent target analytes.