1Inorganic Materials Reserach Program, School of Physical and Chemical Sciences QUT, Brisbane, Australia,
2The Key Laboratory of food Science of Ministry of Education, Nanchang University, Nanchang, China,
3Department of Chemistry,Nanchang University, Nanchang, China
Methods for classification and detection of adulteration of oils are required for quality assurance. Many instrumental techniques and methods have been investigated and used, but often very simple traditional physico-chemical ratings are employed such as colour, moisture, density, refractive index, acid number, saponification and peroxide values. Although relatively simple and inexpensive to perform, the multivariate potential of such measurements for oil characterization have not been explored extensively.
In this work, a number of different chemometrics methods have been employed to characterise and discriminate pure and rancid soybean and rapeseed oils. The seven oil quality parameters were measured for sixty one samples of the two oils in pure and rancid states. The discrimination and characterization of the oils were carried out by partial least squares, artificial neural networks – radial basis function, and PROMETHEE and GAIA multicriteria decision making methods. The performance of these models is compared.