MARÍA J. MARTELO-VIDAL, MANUEL VÁZQUEZ*
*Corresponding author
University of Santiago de Compostela, Faculty of Veterinary Science,
Department of Analytical Chemistry, 27002 Lugo, Spain

Abstract

Artificial neural networks (ANN) combined with ultraviolet (UV), visible (VIS) and near-infrared (NIR) spectral analysis were evaluated as a rapid method to classify wines by grape varieties. Godello, treixadura, albariño and palomino varieties were classified. Results showed that wines from godello, treixadura and palomino grapes were 100 percent classified with the ANN developed in this study. Albariño wines had the worst classification (96.77 percent). It was confirmed the feasibility of applying ANN and UV-VIS-NIR analysis to the authentication of grape variety of white wines.

INTRODUCTION

Wine contains mainly a mixture of water, alcohols and a great variety of organic and inorganic substances such as sugars, polyphenols, amino acids and minerals from the grape juices or produced by the grape fermentation (1). Wine composition depends of soil, culture conditions, microclimate, macroclimate and winemaking techniques (2, 3).
Quantitative and qualitative analysis in wine is not an easy task due to heterogeneity and complexity of wine. The economic value of wine made the wine authentication or classification an important task worldwide. Many Regulatory Councils of designation of origin (DO) are interesting in this task. They have also the difficult task to determine the grape varieties used. This importance is related to quality, prevention of adulterations, food safety and control of winemaking process. Wine quality in import-export markets should be also guaranteed (4). Winemakers and consumers demand analytical low-cost and effective tools to determine the quality of wine (5).
Many analytical techniques have been used to obtain quality control of wine. These include high-performance liquid chromatography (6) or ...