Abstract: The discriminant models of adulterated milk and pure milk were constructed using two-dimensional (2D) infrared correlation spectroscopy by PARAFAC and multivariable linear regression (MLR). First, a total of 96 samples including 48 pure milk samples and three types of adulterated milk (16 melamine-tainted milk, 16 urea-tainted milk, and 16 tetracycline-tainted milk) were prepared. The concentration ranges of all adulterants were 0.01~0.30 g•L^-1. The mid-infrared spectra of all samples were measured in the regions of 900~1 700 cm^-1. Then, the synchronous 2D correlation spectra of all samples were calculated in the region between 900~1 200 cm^-1 and 1 200~1 700 cm^-1. The 2D correlation spectra of all samples were analyzed based on trilinear decomposition using PARAFAC. Finally, the discriminant models for melamine-tainted milk, urea-tainted milk and tetracycline-tainted milk were constructed combined score matrix extracted from 2D correlation spectra using PARAFAC with MLR. The unknown samples were predicted using the constructed models in prediction set. The results show that using a combination of 2D IR correlation spectra and PARAFAC-MLR is an effective analytical method for the classification of adulterated milk and pure milk.

Key words: two-dimensional infrared correlation spectroscopy, PARAFAC, multivariable linear regression, adulterated milk