Desenvolvimento de métodos analíticos para cachaças utilizando espectrofluorimetria e modelos quimiométricos

Abstract

This thesis aimed at the application of chemometric methods in combination with spectrofluorimetry for the analysis of distilled beverages, especially cachaças, contributing to improve the quality control of such symbolic Brazilian beverage. In the first application, a supervised classification model for the discrimination between cachaça samples aged in oak and rum using the synchronous fluorescence spectra was developed. The spectra were preprocessed employing the first derivative with the Savitzky-Golay smoothing, with the parameters (namely, filter width and polynomial order) optimized using a face-centered central composite design. The best suited partial least squares - discriminant analysis (PLS-DA) model was obtained using the synchronous spectra recorded at Δλ = 10 nm, which was able to discriminate the two classes of beverages with an efficiency of 98%, sensitivity of 100 and 94%, and specificity of 100 and 92%, in the training and test sets, respectively, and low false-positive and false-negative rates. In the second application, a method for the classification of cachaças aged in different woods (namely, bálsamo, oak, jequitibá, and umburana) using the synchronous fluorescence spectra, combined with PLS-DA, of commercial samples aged and stored in barrels was proposed. The best suited model was built using the spectra recorded at Δλ = 30 nm, showing efficiency rates ranging from 91 to 100%, sensitivity from 81 to 100%, and specificity from 91 to 100%, in the training and test sets. Commercial cachaças were classified as stored and aged using synchronous fluorescence spectra in the third (and last) application. The best PLS-DA model was obtained using the spectra recorded at Δλ = 20 nm, reaching sensitivity rates of 86 and 81%, specificity of 77 and 71%, and efficiency of 80 and 75%, in the training and test sets, respectively.