Olive oil characterization using excitation-emission fluorescence spectroscopy and three-way methods of analysis

Identificador:  TDX:778

Handle: https://hdl.handle.net/20.500.11797/TDX778

Autores:  Guimet Vila, Francesca

Resumen:
Olive oil contains fluorescent species, such as vitamin E, chlorophylls and oxidation products produced due to degradation. For this reason, fluorescence measurements can be used to obtain information about oil composition. This information can be obtained from direct measurements without needing any prior step of dilution of addition of reagents. This has great advantages with respect other techniques that require these steps, such as chromatographic, because no additional residues are generated and time of analysis is reduced. However the possibilities of fluorescence applied to olive oil analysis are not well studied and this technique is not normally used by the laboratories of olive oil quality control. The current instrumentation allows recording a set of fluorescence spectra at several excitation wavelengths in one measure. As a result, a fluorescence excitation-emission matrix is obtained. Mathematically, fluorescence matrices are matrices of numbers and are called second-order data. Chemometric methods applied to second-order data are called three-way methods, because the data are arranged in a three-dimensional structure. The objective of this thesis has been to characterize olive oils on the basis of excitation-emission fluorescence spectroscopy (EEFS) and three-way methods.The first part of the thesis includes the application of three-way exploratory methods of analysis to a set of commercial oils (virgin, pure and olive-pomace). The methods applied were unfold principal component analysis (unfold-PCA), parallel factor analysis (PARAFAC) and hierarchical cluster analysis (HCA). These methods enabled to distinguish between the oil types on the basis of their fluorescence matrices. PARAFAC has the advantage of providing the underlying spectra of the main families of fluorescent compounds.Next a second set of oils was considered for relating their fluorescence matrices with two quality parameters: peroxide value and K270. The study involved comparing two chemometric methods: PARAFAC combined with multiple lineal regression and multi-way partial least squares regression (N-PLS). N-PLS provided the best fits and the lowest prediction errors. It was also observed that EEFS can detect degraded virgin oils.The last part of the thesis contains the application of several three-way classification methods for detecting adulterations and discriminating between oils on the basis of their origin and type. Firstly, we studied the detection of extra virgin olive oil adulterations with olive-pomace oils (at 5% level). We applied Hotelling T2 and Q statistics and the methods Fisher's linear discriminant analysis and discriminant N-PLS. Discriminant N-PLS differentiated between virgin and adulterated oils with a 100% of correct classification. Next, it is shown that the different stability of the oils from the two subregions of the protected denomination of origin 'Siurana' can be detected from their fluorescence matrices. The method discriminant unfold-PLS enabled to discriminate between the oils from the regions 'Camp de Tarragona' and 'Montsant' with classifications above 90%.Finally, we applied the algorithm non-negative matrix factorization (NMF) to the fluorescence matrices of three sets of olive oils. In this study it is shown that this method is able to decompose the fluorescence matrices of oils into positive parts that can be related with some fluorescence species of oils. We also studied the possibilities of NMF combined with Fisher's LDA for olive oil classification. We obtained classifications between 90-100%.