Abstract: Here is a new generalized rank annihilation strategy by which the application of principal Component analysis to the calibration and prediction samples is performed. The number of principal components are determined according to the percentage of a eigenvalue in the total sum of eigenvalues acquired in the singular-value decomposition. This strategy was used to quantitatively analyze the multinuclear aromatic hydrocarbon mixtures consisting of naphthalene, phenanthrene, pyrene and fluorene by fluorimetry. The influence of the number of pricipal components is discussed and the averaged percent error of prediction is <8% and the largest error is about 13%.

Key words: Generalized rank annihilation, Condensed nucleus aromatic hydrocarbon, Fiuorimetry