An electronic tongue (ET) comprising 18 potentiometric chemical sensors was applied to the quantitative analysis of beer. Fifty Belgian and Dutch beers of different types were measured using the ET. The same samples were analyzed using conventional analytical techniques with respect to the main physicochemical parameters. Only non-correlated physicochemical parameters were retained for further analysis, which were real extract, real fermentation degree, alcohol content, pH. bitterness, color, polyphenol and CO2 content. Relationship between the ET and physicochemical datasets was studied using Canonical Correlation Analysis (CCA). Four significant canonical variates were extracted using CCA. Correlation was observed between 6 physicochemical variables (real extract and fermentation degree, bitterness, pH, alcohol and polyphenols' content) and 14 sensors of the ET. The feasibility of the ET for the quantification of bitterness in beer was evaluated in the aqueous solutions of isomerised hop extract and in the set of 11 beers with bitterness varying between 14 and 38 EBU (European Bitterness Units). Sensors displayed good sensitivity to isomerised hop extract and good prediction of the bitterness in beer was obtained. Calibration models with respect to the physicochemical parameters using ET measurements in 50 Belgian and Dutch beer samples were calculated by Partial Least Square regression. The ET was capable of predicting such parameters as real extract, alcohol and polyphenol content and bitterness, the latter with Root Mean Square Error of Prediction (RMSEP) of 2.5. (C) 2009 Elsevier B.V. All rights reserved.
