© 2017 Elsevier B.V. This work reports the applicability of a voltammetric sensor array able to evaluate the content of the metabolites of the Brett defect: 4-ethylphenol, 4-ethylguaiacol and 4-ethylcatechol in spiked wine samples using the electronic tongue (ET) principles. The ET used cyclic voltammetry signals, obtained from an array of six graphite epoxy modified composite electrodes; these were compressed using Discrete Wavelet transform while chemometric tools, among these artificial neural networks (ANNs), were employed to build the quantitative prediction model. In this manner, a set of standards based on a modified full factorial design and ranging from 0 to 25 mg L−1 on each phenol, was prepared to build the model; afterwards, the model was validated with an external test set. The model successfully predicted the concentration of the three considered phenols with a normalized root mean square error of 0.02 and 0.05, for the training and test subsets respectively, and correlation coefficients better than 0.958.
|Publication status||Published - 1 Mar 2018|
- Artificial neural networks
- Brettanomyces defect
- Electronic tongue
- Phenolic defects