Talanta 64 (2004) 778–784 Monitoring large scale wine fermentations with infrared spectroscopy Alejandra Urtubia a , J. Ricardo Pérez-Correa a,∗ , Marc Meurens b , Eduardo Agosin a a Departamento de Ingenier´ ıa Qu´ ımica y Bioprocesos, Escuela de Ingenier´ ıa, Pontificia Universidad Católica de Chile, Vicuña Mackenna 4860, Casilla 306, Santiago 22, Chile b Unité de Biochimie de la Nutrition, Faculté d’Ingénierie Biologique, Agronomique et Environnementale, Université Catholique de Louvain, Bˆ atiment Kellner, local d231 Croix du Sud, 2(8) 1348 Louvain-la Neuve, Belgium Received 7 December 2003; received in revised form 2 April 2004; accepted 8 April 2004 Available online 1 June 2004 Abstract Negative effects on wine quality and productivity caused by stuck and sluggish fermentations can be reduced significantly, if such problems are detected early through periodic chemical analysis. Infrared spectroscopy (IR) has been used successfully for monitoring fermentations, since many compounds can be measured quickly from a single sample without prior treatment. Nevertheless, few applications of this technology in large scale winemaking have been reported, and these do not cover the entire fermentation from must to finished wine. In this work, we developed IR calibrations for analyzing the fermenting must at any stage of fermentation. The calibration model was obtained with multivariable partial least squares and proved effective for analyzing Cabernet Sauvignon fermentations for glucose, fructose, glycerol, ethanol, and the organic acids; malic, tartaric, succinic, lactic, acetic, and citric. Upon external validation we found an average relative predictive error of 4.8%. Malic acid showed the largest relative predictive error (8.7%). In addition, external validation found that insufficient data for these calibrations made the analysis of fermenting musts using other grape varieties less reliable. © 2004 Elsevier B.V. All rights reserved. Keywords: Cabernet sauvignon; Winemaking; Sugars; Organic acids; Glycerol; PLS 1. Introduction In a highly competitive market wineries need to invest more in technology, to increase productivity and improve average quality, to remain competitive. By reducing stuck and sluggish fermentations, which in turn requires a mon- itoring system that can detect and classify them early, less wine will be lost or downgraded. Artificial intelli- gence techniques have been applied to predict the result of other fermentation processes early on (Kamimura [1,2]; Stephanopoulos [3]) by analyzing significant variables over time. Several compounds play a key role in problematic wine fermentations, such as sugars, nitrogen substrates and organic acids (Bisson [4,5]; Boulton [6]; Pszczólkowski [7]). As conventional chemical analysis is both time con- ∗ Corresponding author. Tel.: +56-2-3544258; fax: +56-2-3545803. E-mail address: pereza@ing.puc.cl (J. Ricardo P´ erez-Correa). suming and expensive, currently these compounds are not measured frequently enough. This problem is more acute in large wineries that operate hundreds of fermentation tanks simultaneously. Infrared spectroscopy offers an alternative to conven- tional chemical analysis. This analytical technique has been applied successfully in other kinds of bioprocess such as the production of antibiotics and the cultivation of mammalian cells. In these processes infrared spectroscopy has been used for monitoring alanine, glucose, glutamine, leucine, lactate and ammonium (Riley [8–10]; Rhiel [11]; Vaccari [12]; Vaidyanathan [13,14]). Although, we have found many applications of IR to wine analysis, such as for controlling denominations of origin, monitoring wines during the aging process (Palma [15]), classification of red-wine dried-extracts according to their geographic origin (Picque [16]) and discrimination among red wines based on the analysis of their phenolic extracts (Edelman [17]), 0039-9140/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2004.04.005