Chemometric characterization of Italian wines by thin-film multisensors array and artificial neural networks

In the present work, nine samples of Italian wines (three white, three red and three rosè) from different denominations of origin have been analysed by the static headspace sampling method to attempt to classify them by chemometric characterization of the data obtained from a thin-film multisensor array. All wines have also been analysed to measure their ionic conductivity, pH and alcoholic content. An electronic nose comprising four metal oxide semiconductor thin-film sensors has been used to generate a typical chemical fingerprint (pattern) of the volatile compounds present in the wines. Principal component analysis and artificial neural networks were applied to the generated patterns to achieve various classification tasks. The classification performance of nine different pre-processing algorithms has been studied on the basis of three different sensor parameters and three different normalization techniques. The wine patterns generation with array sensor signals and the chemometric treatment are fast and simple by providing a recognition rate and a prediction rate as fairly high as 100% and 78%, respectively. These results can be considered satisfactory and acceptable, with the selected variables useful to differentiate these wines by their class.