Quantificação de teores de azeite de oliva em misturas binárias e ternárias, por espectroscopia de infravermelho médio com uso de calibração multivariada

Abstract

Olive oil is a man-made food more than 3000 a. C, and its added value makes it the target of tampering with other vegetable oils. The infrared technique with total attenuated and transformed Fourier reflectance (ATR-FTIR) is a fast and non-destructive medium for plant oil analysis, and requires the minimum preparation of samples, taking speed in obtaining the spectra and good resolution Spectral (4 cm-1). For the reason of adulteration of olive oil, it becomes necessary the development of analytical methodologies. Thus the work aimed to develop a methodology for quantification of olive oil in mixture ternary with soy and canola oils, and binary mixture with sunflower oil. The spectra of vegetable oil mixtures were obtained in the infrared spectrometer with a total attenuated reflectance module. The spectra were normalized and subjected to the multiplicative scatter correction (MSC), being mean centered or subjected to first and second derivatives. The spectra of the samples and your respectively concentrations of olive oil were divided into calibration sets and validation by the Kennard-Stone algorithm. For the selection of variables used in multi-linear regression calibration (RLM), The algorithm of successive projections (SPA) was used with MATLAB software. The square root values of the quadratic medium error (RMSE) for the RLM models without processing and with MSC correction ranged from 2.84 to 3.03, and 2.31 to 2.86, respectively. In the RLM models obtained from the first and second derivatives the RMSE was around 3.26 to 4,19, and 7.09 to 7.24, respectively. The values of the detection limits were in the range of 0.13 g 100 g-1 to 5, 53g 100 g-1, and the quantification limits ranged from 0.43 g 100g-1 to 18.25 g 100 g-1. The accuracy was calculated from the average quadratic prediction error (RMSEP), with results for the models of olive, soy and canola mixtures in the range of 2.84 to 7.15, and for the models of olive oil and sunflower oil mixtures in the range from 0.18 to 1.61. From the developed method it is possible to determine the olive oil content within the calibration range (0 to 100%) for the mixtures of olive oil with soy oils, canola, and sunflower, with good predictive capacity.