Use of Statistical Methods for the Interpretation of Peruvian Wines Measurement Results with MOS Electronic Noses

Abstract

The use of statistical methods for the treatment of information from a data set has been used to be able to better interpret the results obtained in the analysis of different samples. An example, the most common, of these statistical methods used is the Principal Component Analysis (PCA). But, other techniques like Cluster and Factorial analysis allow to have too a better idea of the results.The tin oxides palladium or platinum doped were first prepared by the wet impregnation method. The sensors were made with the screen-printing method to print gold electrodes on alumina plates and finally the oxides, in a mixed with terpineol and ethyl celulose, were put on the electrodes surface. Then, some oxides were modified with a coating of Zeolite Y.With the sensors made in the laboratory, two electronic noses were built. One of them, of platinum doped tin oxides and the other of Palladium (with and without Zeolite Y coating). The sensing of the wines was monitored with the LabVIEW software and the statistical methods were applied with the XLSTAT software. It was observed that for the nose of platinum sensor a difference in the groups of commercial and unknown wines was shown. This difference is attributed to the brand or origin of the wines. While the nose of palladium sensors allowed to distinguish the strains because it locked the Borgoña wines in a different conglomerate. This could be observed with the PCA, cluster and Factorial methods, complementing the three methods the statistical information.The cluster method allowed classifying the results obtained with the detection of Peruvian wines, both when using palladium- and platinum-doped oxides. In both cases, it was possible to differentiate commercial wines from those of unknown brand. In addition, in each case one of the wines was distinguished by some characteristic. For example, the DM (Dulce Mistela) is different than the other wines of unknown brand since these other three are Borgoña. It was observed that the distances between classes are optimal when platinum-doped sensors are used, since the interclass variation exceeds 80%.It is important to note that the results obtained with the factorial analysis mainly reinforce and complement the Cluster analysis, with which very similar groups were obtained, if not practically the same.For example, using Factorial analysis, it is clearly observed that E-Nose 1 (Electronic nose made up of tin oxide-based sensors doped with different concentrations of platinum (0.1; 0.2; 0.3 and 0.5% Pt) with and without zeolite coating, and with three additional commercial sensors) separates the wines into two groups that are well differentiated: the known commercial wine (blue circle) from the unknown (red circle). It is also observed that the variance value of the two factors is high 99.26%. That could be seen in Figure 1. Figure 1