Desenvolvimento de nariz eletrônico de baixo custo: comparativo entre a discriminação instrumental e perfil descritivo quantitativo na análise de cafés torrados comerciais

• Ano de defesa: 2018
• Autor(a) principal: Makimori, Gustavo Yasuo Figueiredo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso embargado
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Campo Mourao; Medianeira; Brasil; Programa de Pós-Graduação em Tecnologia de Alimentos; UTFPR
• Palavras-chave em Português: Avaliação sensorial; Quimiometria; Arduino (Controlador programável); Sensory evaluation; Chemometrics; Arduino (Programmable controller); CNPQ::CIENCIAS AGRARIAS::CIENCIA E TECNOLOGIA DE ALIMENTOS; Tecnologia de Alimentos
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/3955

Coffee is a beverage with an extremely complex aroma and flavor. Factors such as the geographic region, climatic conditions, and type of management, roasting methods, milling and storage are intrinsically related to the quality and characteristics of the final product. One of the ways to evaluate coffee quality is through quantitative descriptive profile (QDP), which is a sensory analysis that is based on the training of selected assessors that evaluates the intensity of sample attributes. The QDP presents some disadvantages such as long periods of analysis, cost and high consumption of samples. Thus, rapid instrumental methods allied to chemometrics methods could be used to control the quality of coffee to replace or at least reduce the amount of samples analyzed by traditional sensorial methods. Electronic nose (e-nose) are instruments based on an array of partially selective chemical sensors capable of detecting volatile organic compounds. With the high cost of commercial e-nose and the complexity of the data collected the development of prototypes with low cost components and the application of recent chemometric methods is an opportunity. On the other hand, spectra in the visible region (Vis) can inform the coffee roast degree and the near infrared (NIR) provide information on the overtones of the molecules vibrations and are widely used to identify chemical characteristics due to quick analysis and preservation of the sample. In addition, the electronic tongue (e-tongue), which is composed of a non-specific, low-selectivity chemical multi-sensor system, can also be used to evaluate coffee. The objective of this work was to develop an e-nose prototype and evaluate 12 commercial roasted and ground coffee samples by the three fast instrumental methods already mentioned (e-nose, e-tongue and Vis-NIR spectra) to verify if they are capable of discriminating different commercial coffees in a similar way as QDP. The e-nose prototype was built from a development board (Arduino Mega 2560) and MOS-type sensors (metal-oxide semiconductors) that detect volatile organic compounds (VOCs). Control algorithms, automation and graphic interface were developed in MATLAB® environment. The signal obtained by the e-nose lab-made showed good repeatability when tested by means of a ternary mixture of the simplex centroid type (acetone, acetonitrile and ethanol). QDP was conducted with 15 assessors trained for two months. Six attributes of odor, three of flavor, one of texture (body of the beverage) and color were used to characterize the products. The QDP data were analyzed in two different ways. The first one followed a commonly used strategy, Analysis of Variance (ANOVA), Tukey's Test, Principal Component Analysis (PCA) and spider chart. The second strategy was to use the Common Dimension Analysis (ComDim) in both the selection of attributes and assessors as well as the final analysis of the PDQ data. Even by selecting different assessors, both techniques described the 12 products in a similar way. However, with ComDim the interpretation is facilitated by the graphs of scores and loadings. Products containing 100% of arabica coffee were described with high intensity by green odor and fresh odor attributes, since the blends products were described with high intensity attributes of bitter taste, body of the beverage and brown color in the two methods. Both strategies emphasized that the products "A" and "B" are the most different samples of the set and the color was the attribute used more efficiently by the assessors to distinguish the products. Finally, for the twelve coffees, the Vis-NIR spectra, the e-nose signal, the e-tongue signal and the QDP attributes were combined in an exploratory multiblock analysis using the ComDim. With 51.96% of the total variance, the common dimension 1 (CD 1) has high saliences value for the five techniques used, that is, they all carry a common information in their data. However, instrumental techniques also have a high saliences value in other dimensions, Vis CD 2, e-nose CD 3, e-tongue CD 4 and NIR CD 5, that is, each instrumental technique used carries a complementary information to the QDP that it possesses high salience only in CD 1. Therefore, when the information of the instrumental techniques used are combined, the discrimination of the products resembles the result of the QDP. Coffee A remained as the most different for all instrumental methods. Coffees D, G and J showed similarity in all instrumental methods, always remaining in the same quadrant. It was also possible to verify the influence of the degree of roast and the robusta coffee content in the blends. Therefore, the development of fast and low-cost instrumental methodologies combined with chemometric methods provide an alternative perspective for the analysis of coffee and even of other food matrices.