Otimização das análises de leite para detecção de soro por espectroscopia de infravermelho

Abstract

The dairy production chain is of great economic and social importance to Brazil, with the country being the third largest producer of milk in the world. The addition of whey to the milk is one of the most common frauds, jeopardizing the quality and nutritional value of the food. Fraud detection has been carried out using traditional analytical methods, but electronic equipment, such as infrared spectroscopy, is also being used to facilitate the evaluation of milk quality. The aim of this study was to develop a predictive and/or classifying model to detect the addition of whey in quantities that would allow the milk to be classified according to Normative Instruction N°68. Evaluating the efficiency of the near-infrared spectroscopy method in the quantification/detection of whey in raw milk and comparing it with the method used in routine use, the c-Gmp Kit. In the study, samples of raw milk were collected during the milk reception routine at the Nestlé factory in Montes Claros, Minas Gerais. The samples were divided into two groups: blank samples, which had no adulteration with serum, and adulterated samples, which had intentional addition of serum. A total of 886 samples of the routes received at the unit between January 2022 and August 2023 were collected. To check for intentional contamination during the study, analyzes of the CMP index of 5 levels of the concentrations used (0 - free of addition, 30 mg/L, 50 mg/L, 50 mg/L, 70 mg/L, 100 mg/L and 150 mg/L) were performed. The checks were performed by the HPLC method as well as by the c-GMP kit. The method showed a good correlation with the HPLC methodology. However, when comparing the method with the c-GMP kit, it was observed that the equipment was not able to quantify satisfactorily at levels between 30 mg/L and 75 mg/L. However, the equipment demonstrated that when taking readings at levels above 75 mg/L, it also quantifies at lower levels (between 30 mg/L and 75 mg/L), which may be useful in quantifying levels above 75 mg/L. To reduce the error found in the detection method, it will be necessary to carry out a more in-depth study and to develop the use of other tools, such as Artificial Neural Networks (RNA).