Analytical and chemometric methodologies for fruit agroindustry
NIR. Chemometric. Fruit. Jabuticabas. Castanholas. Classification. Regression
This thesis was developed with the aim of contributing to the repertoire of analytical techniques for the fruit agroindustry, using the near infrared, through a chemometric treatment. The application of methodologies through non-destructive techniques, such as the near-infrared, to the fruit agroindustry is justified by the growing demand for products and the need for technologies that can accompany the growth of this market, with the advantages of reducing analysis time, costs for reagents, work for the analyst and generation of waste in the analyzes. Two types of approaches, one application for fruit classification, and the other for the quantification of sensorial and nutritional quality evaluation parameters of fruits were carried out. For fruit classification, jabuticaba were analyzed in three stages of maturation (immature fruit, physiologically mature and mature) in the near infrared. Then, models were developed for the classification of these three stages of maturation, using the algorithms PCA-LDA, SPA-LDA and GA-LDA, being the best result found when GA-LDA was used. class separation. The results demonstrated the sensitivity and the ability of the method to evaluate the maturation of jabuticaba, as an indication of the possibility for its industrial application. In the approach of quantification of parameters in fruits, the content of anthocyanins and content of phenolic compounds in castanets by NIR were determined. After that, regression models were developed to predict these parameters, using PLS and variable selection using iPLS-PLS and GA-PLS. The best results found for the determination of phenolic compounds and anthocyanins were with GA-PLS for both parameters. The best model found for phenolic compounds presented a prediction correlation (𝑅𝑝2) of 0.82 and a prediction error (RMSEP) of 11.3 mg GAE g-1 (mg equivalents of gallic acid (GAE) per g sample. (𝑅𝑝2) of 0.80 and a predictive error (RMSEP) of 8.7 mg L-1. In view of the results obtained, it is possible to conclude that the allied NIR chemometric methods presents with potential applicability to analytical methodologies in the fruit agroindustry, both for qualitative as well as quantitative approaches.