Exposing adulteration of Muscatel wines and assessing its distribution chain with fluorescence via intelligent and chaotic networks

Abstract

In this work, intelligent algorithms and fluorescent measurements have been integrated into a portable system to evaluate the storage conditions and detect adulterations of wines such as water and ethanol. More than 700 spectra derived from the analysis of three types of either large-scale or artisanal Muscatel wines of European origin (French and Russian) were collected. Different sets of independent variables were extracted and used to train neural network models. Two classes were employed including variables extracted via feature selection directly from the fluorescent emission spectra and others calculated in the form of chaotic parameters. To reach the two proposed objectives, more than 77,500 neural networks have been developed to optimize the tools. The main results are that the integration between fluorescence and intelligent algorithms, whether based on chaotic parameters or direct emission data, are capable of detecting anomalous storage conditions and, above all, of locating the presence of adulterants such as water or ethanol in the wines tested.