Synchronous fluorescence spectroscopy and multivariate classification for the discrimination of cachaças and rums

Although cachaça and rum are distilled beverages obtained from the same raw material, they present differences in their chemical compositions. In this study, synchronous fluorescence spectroscopy was used combined with supervised classification models based on the partial least squares discriminant analysis to develop a rapid and low-cost model for discriminating between 50 cachaça and 40 rum samples. Partial least squares discriminant analysis models were constructed using synchronous fluorescence spectra recorded at wavelength differences of 10-100 nm. Initially, spectra were preprocessed by the first derivative with the Savitzky-Golay smoothing, and filter width and polynomial order were selected through face-centered central composite designs. For the construction and validation models, the spectra data were split into two datasets: the training and the test sets containing 60 (C, n = 33; R, n = 27) and 30 (C, n = 17; R, n = 13) samples, respectively. The best discrimination was achieved using fluorescence spectra recorded at wavelength difference 10 nm, allowing the discrimination of cachaça and rum with a classification efficiency of 98%. These results indicate that synchronous fluorescence spectroscopy offers a promising approach for the authentication of cachaças and rums.