A new procedure, bsed on several recent chemometric method, is proposed for automatic analysis of data from evolutionary process. It runs in the following steps: first, select key spectra or variables; second, determine the chemical rank of data by subspace coparison; third, iterative into pure spectra and concentration profiles by elementary transform. Criteria, e.g. non-negativity, are used to check the reliability of results. As far as data from GC-MS is concerned, key variables can also he used for analysis besides key spectra. The proposed procedure greatly reduces user intervention and therefore saves analysis time. It promises a powerful tool for the resolution of complex system. There are 557 components resolved from petroleum sample in this paper.

Key words: CHROMATOGRAPHY-MASS SPECTROGRAPHY, PETROLEUM, AUTOMATIC ANALYSIS, STOICHIOMETRY