Slice spectra approach to synchronous Two-dimensional correlation spectroscopy analysis for milk adulteration discriminate

The discrimination approach of adulterated milk was proposed combined synchronous two-trace twodimensional (2T2D) correlation slice spectra at the characteristic wavebands of adulterant in milk with multivariate method. Two common adulterants, melamine and urea, were analyzed to demonstrate useful by the method. 2T2D (near infrared) NIR slice spectra at characteristic wavebands of adulterant were extracted from the synchronous 2T2D correlation spectra, and were input to construct the N-way partial least squares discriminant analysis (NPLS-DA) models. One-dimensional (1D) spectroscopy featuring all the present components in the samples combined with partial least squares discriminant analysis (PLSDA) was also evaluated for comparison. The results indicated that for one kind of adulterant in model, prediction accuracies of slice spectral models were both 100% for melamine-adulterated and ureaadulterated samples discrimination. Moreover, for two kinds of adulterants in model, prediction accuracies of slice spectral models were 90.57% and 100% for melamine-adulterated and urea-adulterated discrimination, respectively, which was better than those of 1D whole models based on PLS-DA (only 81.13% and 98.15%, respectively). The comparison informs that the 2T2D slice spectra extracted at the characteristic wavebands of adulterant highlighted the adulterant spectral features and was obviously advantage to improve the discrimination accuracy. Meanwhile, the complexity of slice spectra is significantly reduced compared with the whole matrix of synchronous 2T2D correlation spectra. (c) 2022 Elsevier B.V. All rights reserved.

Automated summary

The discrimination approach of adulterated milk was proposed using synchronous two-trace two-dimensional (2T2D) correlation slice spectra and multivariate method. The method was demonstrated to be useful for analyzing two common adulterants, melamine and urea. The study showed that the prediction accuracies of the slice spectral models were higher than those of the 1D whole models, indicating that the 2T2D slice spectra improved the discrimination accuracy of adulterants in milk. The complexity of the slice spectra was significantly reduced compared to the whole matrix of synchronous 2T2D correlation spectra, making it a more efficient method for adulterated milk detection.