Aplicação de técnicas espectroscópicas e métodos de modelagem de classe na discriminação geográfica de grãos de café verde da região do Cerrado Mineiro

• Ano de defesa: 2022
• Autor(a) principal: Larissa Batista dos Santos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ICX - DEPARTAMENTO DE QUÍMICA; Programa de Pós-Graduação em Química; UFMG
• Palavras-chave em Português: Café; Denominação de origem; Modelagem independente e flexível por analogia de classe SIMCA; Poder de modelagem; Fusão de dados; Soft independent modelling by class analogy; Coffee; Protect designation of origin; Data fusion; Modelling power; Química analítica; Minas Gerais; Certificados de procedência; Espectroscopia de infravermelho; Espectrometria de massa; Espectroscopia de absorção atômica; Fluorescência de raio X; Mínimos quadrados; Modelagem de dados
• Link de acesso: http://hdl.handle.net/1843/45077

Coffee is one of the most consumed, appreciated beverages in the world. In the economic context, the raw material is widely relevant for Brazil, especially Minas Gerais. With the technological advance and improvement in the life’s quality, the search for products or services that have differential is increasing. The green coffee beans produced in the Cerrado Mineiro region has Protected Designation of Origin (PDO) certificate that guarantees the quality and differential of the beans. In this context, the objective of this work was to develop classification models to characterize coffee beans from the Cerrado Mineiro. Spectroscopic techniques, total reflection X-ray fluorescence (TXRF), attenuated total reflectance mid-infrared spectroscopy (ATR-MIR), paper-spray mass spectrometry (PS-MS) and ultraviolet-absorption spectroscopy (UV-Vis) were used in this work. Design of experiments were constructed to optimize the extraction of compounds present in green coffee beans to be used in PS-MS and UV-Vis analyses. Class-modelling methods SIMCA (Soft Independent Modelling by Class Analogy), DD-SIMCA (Data driven Soft Independent Modelling by Class Analogy) and OCPLS (One class partial least squares) were built for the individual data block of each technique and with the concatenated data to take advantage of the synergy between the data from different techniques. The OPS variable selection method was applied to improve the performance of the model. In general, models built with UV-Vis data and data fusion from the other techniques performed better. The variable selection method was able to select the most important variables for the models, aiming to improve their performance. The interpretation of the models was carried out through the modeling power of the variables in which it was possible to observe that trigonelline and chlorogenic acids substances were responsible for the discrimination of coffee beans from the Cerrado region in relation to coffee beans from the Caparaó, Mogiana and South of Minas region. Regarding the inorganic elements, P, Cl, Ti, Cu, Zn and Rb were selected as the most important variables from this dataset. The performance of the models was interpreted estimating the figures of merit, sensitivity, specificity, and efficiency.