Application of E-nose combined with ANN modelling for qualitative and quantitative analysis of benzoic acid in cola-type beverages

Effective detection methods are of critical importance to food surveillance for preservatives in commercial food and beverages, aiding to avoid excessive intake by consumers. In this study, E-nose technology was combined with ANN modelling for qualitative and quantitative analysis of benzoic acid in cola-type carbonated beverages. The qualitative model generated accuracy rates of 90.0 and 92.0% in category identification for test samples with High, Medium or Low levels of benzoic acid, when performed on the testing and validating data subsets, respectively. For quantitative analysis, the predicted values of benzoic acid exhibited strong linear correlation with reference values determined with HPLC, with R2 between predicted and reference values being 0.99, for both the testing and validating data subsets; mean relative differences of predicted values compared with reference values were 4.47 ± 3.66 and 1.93 ± 3.01% for the testing and validating data subsets, respectively. Paired t-test indicated that the differences between predicted and reference values were statistically insignificant (P > 0.05). Results of this study implied that E-nose combined with ANN modelling was reliable to determine benzoic acid in carbonated beverages, and could be utilized as a useful tool for effective surveillance of benzoic acid in commercial beverages in market.