Calibração multivariada de misturas de óleos vegetais utilizando espectroscopia no infravermelho médio

Abstract

Vegetable oils are very important for human health because they are excellent sources of monounsaturated and polyunsaturated fats, which are beneficial to our health. And Brazil is a major producer of vegetable oils, mainly soybean oil, either as raw material for the food industry, whether for the production of biodiesel, where the oil is corresponds to more than 50% of the feedstock that fuel. The technique for determination of many physical and chemical parameters of vegetable oils depended on the use of methods slow and laborious, besides using aggressive reagents to the environment. The Mid Infrared Spectroscopy FT-IR is a technique that has overcome these drawbacks by using analytical procedures quick, direct, accurate, and do not utilize reagents that not only harmful to the environment and human health. Thus, this paper aims, through the methodology of mid-infrared to determine the fatty acids present in mixtures of 4 types of vegetable oils (Peanut, Sesame, Soy and Passion Fruit). The peanut and sesame oils were extracted from seeds of BRS and BRS-Havana-Seda, both supplied by Embrapa Cotton, passion fruit oil was purchased commercially from São Paulo-SP and soybean oil was purchased in trade city of Campina Grande-PB. The mixtures were prepared based on an array of experimental design and then submitted to analysis of fatty acid composition and physico-chemical. Samples were analyzed on a mid-infrared spectrometer Varian Model IR 640, where the properties studied were used as reference variables for the prediction by mid-infrared spectrometry and the data were processed using the partial least squares (PLS) methodology as multivariate calibration. The results obtained by infrared spectroscopy medium underwent regression analysis by partial least squares (PLS), was used for this Uncrambler 9.8 ® software. Prediction models by PLS to fatty acids in the mixtures under study were effective for succinct and direct estimation of the fatty acids present in samples of vegetable oils analyzed. As models for predicting physicochemical properties of the mixtures under study were also effective for the direct estimation of the variables of density, refractive index, viscosity, acid number, iodine and moisture.