Aplicação de calibração multivariada em dados de espectroscopia UV-visível para previsão da acidez total em vinhos

Abstract

This research aimed the study of an alternative methodology to replace acid base titration used to measure total acidity in white and red wines during the industrial process of online measurements. More specifically, this research focused on providing Ultraviolet and Visible (UV-Vis) spectrophotometry as an alternative tool for being itself a methodology with potential to be applied in industry to monitor total acidity since it provide fast analysis without loss in efficiency and without waste generation. In order to apply UV-Vis spectrophotometry it was necessary the multivariate calibration by using Partial Least Squares regression (PLS) between wine UV-Vis spectra and its total acidity values, determined from reference method, which in this case was acid base titration. The best model acquired to predict total acidity in new wine samples was obtained trough chemometric strategies, as Principal Component Analysis (PCA), to observe if the sample presented linear behavior, and Kennard Stone algorithm, used to better select calibration and validation samples. The Root Mean Square Error of Calibration (RMSEC) and Root Mean Square Error of Prediction (RMSEP) values and R2 relating calibration and prediction were evaluated. To the white wines evaluated in the study the values obtained to RMSEC, RMSEP and R2 were 5.87, 6.58 and 0.71, while the model to the red ones provided 6.93; 8.58 and 0.71, respectively. The results suggest that the methodology could be employed since the detection system used in titration is the human eye, being a point of discussion among the errors obtained in it. Moreover, the paired t-test was used to evaluate the compatibility among UV-Vis spectrophotometry and titration, and the results shows that there is no significant difference considering a confidence interval of 95%. In this way, the advantages provided for UV-Vis spectrophotometry are enough attractive to the industries, what suggests that some more studies focusing on optical methods and multivariate calibration could be an interesting issue to the area and deserve to be observed in future studies.