Desenvolvimento de modelo de calibração multivariada multiproduto para quantificação de acidez e vitamina C em néctar de frutas e bebidas industrializadas a base de soja

Abstract

The determination of the total acid and vitamin C using ultraviolet spectroscopy, visible near infrared spectroscopy and multiproduct multivariate calibration are presented as an alternative to the conventional method through potentiometric titration and oxidation-reduction titration, respectively. The model was developed including different products (fruit nectar of different flavors and light fruit nectar, soybean beverages and light soybean beverages). The calibration and validation data sets were optimized through outliers identification. Partial least squares (PLS) regression was used in order to build the models. Accuracy, represented by the Root Mean Squared Errors of Calibration (RMSEC) and Prediction (RMSEP) were determined. The confidence ellipse and adjust confirms that accuracy was achieved, while residues showed random behavior, confirming linearity for the model developed. Figures of merit, such as, sensitivity, analytical sensitivity, residual prediction deviation, limits of detection and quantification were evaluated. The proposed methodology shows advantages as fastness, does not requires sample preparation and does not generate toxic waste, being an alternative to the conventional method based on potentiometry and oxidation-reduction titration.