Desenvolvimento de métodos analíticos para monitoramento da qualidade de farinhas funcionais e cafés

• Ano de defesa: 2018
• Autor(a) principal: Brito, Anna Luiza Bizerra de
• Tipo de documento: Tese
• 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: Alimentos; ICP OES; ICP-MS; Transferência de calibração; NIR; TGA; LDA; Food; Calibration transfer; CNPQ::CIENCIAS EXATAS E DA TERRA::QUIMICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/13045

The search for a better quality of life has led to increased consumption of functional foods. Ensure a quality of food is an aspect of great relevance and, therefore, the parameters of quality in all cases and being constantly monitored by inspection agencies. Thus, this thesis presents the development of analytical methods that were evaluated to monitor some quality parameters in samples of functional flours and coffee. The work was divided into three stages. The first stage (Chapter II) reports the determination of the concentrations of macroelements, microelements and non-essential elements in 60 samples of functional flours of 20 different types. The spectroanalytical techniques: optical emission spectrometry with inductively coupled plasma (ICP OES) and inductively coupled plasma mass spectrometry (ICP-MS) were used in the determinations of these elements. A statistical approach utilizing Principal Component Analysis (PCA) was realized for exploratory data analysis (elements) in characterization of the flour samples. The concentrations of macroelements (K> P> Ca> Mg), ranged from 746.4 up to 30,328 mg kg-1, and concentrations of the microelements (in descending order for Fe, Al, Mn, Cu, Mo, and Cr) ranged from 0.03 up to 160.1 mg kg-1. The non-essential elements, As, Cd and Pb, determined in some samples of the flours presented concentrations that are within the range required for a daily intake. The second stage (Chapter III) were investigates different strategies of calibration model transfer to determine crude protein and fiber properties in samples of flours with functional properties using portable and benchtop near infrared (NIR) region instruments. In the study, direct standardization (DS), Piecewise Direct Standardization (PDS), and reverse standardization (RS) methods were evaluated for the transfer of the spectra obtained with the two instruments and different numbers of transfer samples were tested. The Partial Least Squares with the full spectrum (PLSfull spectrum), PLS with regression coefficients selected by the Jack-Knife algorithm (PLS/JK) and Multiple Linear Regression (MLR) with previous selection of variables by the Successive Projections Algorithm (MLR/SPA) calibration models were evaluated. The results showed that the lowest values of the Root Mean Squared Error of Prediction (RMSEP) were obtained with the PLSfull spectrum (1.10) and MLR/SPA (1.45) models for the crude protein parameter using the DS method, while for fiber the lowest RMSEP value was obtained using the PLSfull spectrum (2.34) with RS, although no statistically significant difference was found among the RMSEP values obtained for the analyzed models, confirmed through F test (95% confidence level). Finally, in the third stage (Chapter IV), Thermogravimetric Analysis (TGA) and Linear Discriminant Analysis (LDA) were used to classify coffee samples as caffeinated or decaffeinated. The classification models were constructed through the association of LDA and algorithms of selection of variables: SPA, Genetic Algorithm (GA), and Stepwise (SW). The GA/LDA model presented no error in the classification of the test samples (caffeinated: 8, decaffeinated: 5), while the SPA/LDA and SW/LDA models presented 1 error in each. The models were compared in terms of accuracy, specificity and sensitivity values obtained for the test subsets, in which GA / LDA presented the best performance with 100% accuracy and 1.00 in sensitivity and specificity values for each class.