Chromatographic pattern recognition for the analysis of complex mixtures

초록

A system of pattern recognition for the analysis of complex mixtures was developed using a conventional capillary gas chromatograph and a personal computer with laboratory-written software. Pattern recognition was accomplished by a multiple reference peak identification method and outlier statistics using an on-line data system. Reference peaks were selected as they were well resolved and present in both the sample and reference chromatograms. The resultant area ratios of peaks selected as the nearest neighbours were treated statistically by three kinds of outlier rejection method, Dixon's Q test, the two-sided Grubbs test and Huber's elimination rule. The outlier peak is simply dropped from the averaging and the reported mean becomes more representative of the true value. Huber's elimination rule gave the best results. Repeatability of the results of analysis was examined by a two-sided analysis of variance (ANOVA) test.

키워드