Classificação de cafés solúveis usando espectroscopia NIR e quimiometria

• Ano de defesa: 2021
• Autor(a) principal: Nóbrega, Rossana Oliveira da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Química; Programa de Pós-Graduação em Química; UFPB
• Palavras-chave em Português: Café; espectroscopia NIR; DD-SIMCA; PLS-DA; iSPA-PLS-DA; Coffee; NIR spectroscopy; CNPQ::CIENCIAS EXATAS E DA TERRA::QUIMICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/21560

Soluble coffee is a drink obtained from the dehydration of roasted coffee extract, the quality of which is characterized by its organoleptic properties, which depend on several factors, such as grain variability, caffeine extraction, grinding, roasting process, among others. However, the quality of this product must be monitored to ensure food safety and acceptance by consumers and regulatory bodies. Thus, this work aimed to develop green analytical methodologies based on bench and portable Near Infrared Spectroscopy (NIR: near infrared spectroscopy) and chemometric to assess the conformity of commercial soluble coffees, regarding the type and degree of roast. The first approach of the study is to classify the samples of soluble decaffeinated coffees in relation to regular soluble coffee (with caffeine), using Data-Driven Independent and Flexible Class Analogy Modeling (DD-SIMCA: Data-Driven – Soft Independent Modeling of Class Analogy) on both bench-top and portable NIR equipment. In this approach, the results obtained were 100% sensitivity, specificity, and accuracy. In the second approach, regular soluble coffees were classified with respect to roasting, traditional and extra-strong, using Partial Least Squares Discriminant Analysis (PLSDA) and the Successive Projections Algorithm for selection of intervals in PLS-DA (iSPA-PLS-DA: Successive Projections Algorithm for Interval Selection in Partial Least-Squares). For bench-top NIR spectra, the best result was obtained with the iSPA-PLS-DA method, when using the moving average pre-processing with multiplicative scatter correction (MM+MSC), reaching 96.7% of accuracy rate in the discrimination of samples in their respective classes. In the case of portable NIR, the best sorting performance was observed for iSPA-PLS-DA with moving average preprocessing and baseline offset correction (MM+BO), with 98% accuracy. Therefore, this study showed the potential of NIR spectroscopy together with chemometric classification tools for rapid, non-destructive, and direct analysis of soluble coffee, which can be useful for evaluating quality parameters during the industrial process, as well as the finished product.