Classificação multivariada de leite UHT quanto à presença de lactose utilizando espectroscopia no infravermelho próximo (NIR)

• Autor(a) principal: LIMA, Glaucia Ferreira de
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFRPE
• Texto Completo: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/7358
• topic: Leite; Lactose; Classificação multivariada; Espectroscopia; CIENCIAS AGRARIAS::CIENCIA E TECNOLOGIA DE ALIMENTOS

Lactose hydrolyzed milk, produced by the conversion of this carbohydrate by galactosidases, was developed since the 1970s mainly to suit individuals with lactose intolerance. However, the analytical methods most used for the analysis of sugars are inefficient for the qualitative evaluation of lactose in products containing different sugars, such as lactose-free milk. Therefore, the use of more sophisticated methods for this analysis becomes essential for the food industry. Near Infrared Spectroscopy (NIRS) represents an emerging analytical technique for the food industry because of characteristics such as non-destructive and rapid analysis. Nevertheless, the use of multivariate data analysis is necessary to extract the encoded information in the spectra. Thus, the aim of this work was to develop and validate a methodology based on NIR spectra obtained in portable and benchtop equipments, associated with the multivariate analysis for the classification of long-life milk regarding the presence of lactose. For this purpose, 71 samples of UHT regular (30) and lactose-free milk (41) were purchased in local commerce. In the present study, the spectra in the diffuse reflectance mode were obtained on high resolution benchtop (FT-IR Spectrum Frontier from Perkin Elmer) and low resolution portable (MicroNIR 1700 from JDSU Corporation) spectrometers. The classification models using the supervised techniques PLS-DA and LDA with variables selection by genetic algorithms (GA) and successive projections (SPA) were developed from the preprocessed spectra. For both equipment, benchtop and portable, all models presented sensitivity and specificity of 100%, only SPA-LDA model for portable equipment obtained sensitivity in 80%. Therefore, NIR spectra, especially those obtained with portable equipment, combined with multivariate classification was effective for the presence identification of lactose in UHT milk.