Wine classification by taste sensors made from ultra-thin films and using neural networks

This paper reports on a sensor array able to distinguish tastes and used to classify red wines. The array comprises sensing units made from Langmuir-Blodgett (LB) films of conducting polymers and lipids and layer-by-layer (LBL) films from chitosan deposited onto gold interdigitated electrodes. Using impedance spectroscopy as the principle of detection, we show that distinct clusters can be identified in principal component analysis (PCA) plots for six types of red wine. Distinction can be made with regard to vintage, vineyard and brands of the red wine. Furthermore, if the data are treated with artificial neural networks (ANNs), this artificial tongue can identify wine samples stored under different conditions. This is illustrated by considering 900 wine samples, obtained with 30 measurements for each of the five bottles of the six wines, which could be recognised with 100% accuracy using the algorithms Standard Backpropagation and Backpropagation momentum in the ANNs. (C) 2003 Elsevier B.V. All rights reserved.