Yoghurt standardization using real-time NIR prediction of milk fat and protein content

Abstract

A system based on near-infrared (NIR) spectroscopy has been developed for the in-line control of the composition of the milk used as raw material for yoghurt production to control the content of protein and fat in the final product, and, therefore, to reduce variability in the production process. Firstly, after selecting the appropriate method for preprocessing NIR data, Partial Least Squares Regression models were built to predict fat and protein content in milk, obtaining good performances. The variance explained of y-block in prediction (R2P) was 0.99 and 0.80, while the Root Mean Square Error of Prediction (RMSEP), was 0.26 and 0.16 for fat and protein, respectively. With those models, it was possible to determine the fat and protein contents in milk in real time, and therefore, the quantity of milk powder and cream added in the manufacturing process of yoghurt could be readjusted. The presented strategy allows the improvement of the homogeneity of the final product, reducing the variability of the nutritional values in more than 70% with respect to the traditional recipe, and also meet the target values according to yoghurt producers for fat and protein content, that is, 10% of fat and 5% of protein.