An improved methodology of soft drink discrimination using an electronic nose

An improved methodology of soft drink discrimination using an electronic nose is developed in this study. 4 kinds of soft drinks, namely Coca Cola, Pepsi Cola, Future Cola and Sprite are detected. 3 pattern recognition techniques, PCA (Principle Component Analysis), MDA (Mahalanobis Distances Analysis) and PNN (Probabilistic Neural Network) are employed to verify the effectiveness of the 3 sampling procedures. The results indicated that, sampling by static headspace, 25 samples are misclassified in PNN analysis. The electronic nose cannot discriminate the 3 Colas due to the presence of humidity in the headspace, only Sprite can be discriminated from the Colas. With 21 samples are misclassified in PNN analysis, the EDU (Enrichment and Desorption Unit) cannot improve the results significantly. Sodium carbonate powder is very effective in adsorbing moisture in the samples, which effectively improves the sensitivity and the stability of the electronic nose sensors. Consequently, all the samples are classified correctly in PNN, and the electronic nose can be used in soft drink detection.