NIR PORTABLE FOR RAPID FAT BLOOM IDENTIFICATION IN CHOCOLATES

One of the main defects that occurs in the chocolate industries is the formation of needle-shaped fat crystals, known as fat bloom. Several factors lead to its formation, but what concerns industries the most are the distribution temperature and inadequate storage conditions. This defect is often evaluated by traditional analytical equipment, but green and fast response techniques such as portable near infrared spectroscopy (NIR) have been used as an alternative. In this context, the objective of this work was to apply portable NIR allied to the partial least squares discriminant analysis (PLS-DA) method to identify the temperature-abused chocolates. Four types of chocolate of the same brand were purchased (white, milk, 40% cocoa and 70% cocoa). These samples were placed at three different temperatures (20°C, 35°C, 40°C). Reflectance spectra were obtained in the range 900 to 1700 nm on an FT-NIR spectrometer (Spec1700 - Speclab). Chemometric analysis was performed using PLS-Toolbox (version 5.8) in Matlab (R2014b). The PLS-DA method was obtained after spectra were pre-processed using 1st derivative (Savitzky-Golay) and mean centering. In general, the models developed to identify the temperature-abused chocolates presented sensitivity and specificity between 86% and 100% for training set and between 50% and 100% for test set, respectively. However, the model that presented the highest classification errors was for milk chocolate, assuming that this occurred due to its composition rich in cocoa butter substitute fats, which aims to increase the thermal stability of the polymorphic structure of chocolate. Therefore, the use of portable NIR proves to be a promising technique for verifying the fat bloom defect, often not observed with the naked eye.