Dados do Trabalho

Título

MINIATURIZED NIR SPECTROMETER AND CHEMOMETRICS: A COMBINATION TO UNMASK ADULTERATIONS IN BRAZILIAN SPECIALTY CANEPHORA COFFEES

Introdução

Authentication of Brazilian specialty Canephora coffees is essential for guaranteeing quality, safety and mitigating fraud. Fast and accurate methods are need to identify coffee adulteration, which can jeopardize consumer confidence and cause negatively impact the market for this high value-added product. In this study, the use of a portable NIR spectrometer combined with Data-Driven Soft Independent Modeling of Class Analogy (DD-SIMCA) is proposed to authenticate specialty Brazilian Canephora coffees against adulterations.

Material e Métodos

Spectrum collection was performed with a MicroNIR (Viavi Solutions Inc., USA) with a resolution of 6.2 nm, from 950 to 1650 nm, and 105 scans, directly from samples placed in a Petri dish. Two groups of samples were analyzed: the first group (authentic samples, n=40) consisted of pure roasted and ground Robusta Amazonian (n=20) and Conilon from the state of Espírito Santo (n=20) coffees. The second group consisted of pure coffees adulterated with spent coffee grounds, low-quality Canephora coffee, coffee husks, corn, and soybeans roasted and ground in increasing proportions of 1, 5, and 10% with two repetitions, totaling 30 adulterated samples. The data were preprocessed with Multiplicative Scatter Correction, First Derivative with Savitzky-Golay smoothing over a 7-point window, and Mean Centering. A Principal Component Analysis (PCA) was performed. For the development of the DD-SIMCA model, data from authentic samples were splited by the Kennard-Stone algorithm into training (n=30) and true positive dataset (n=10).

Resultados e Discussão

The spectra of the adulterated samples constituted the true negative dataset (n=30). The model's quality was evaluated based on sensitivity and specificity values, adopting a significance level of 0.01 for type I and II errors. The PCA explained 99.79% of the variance in two principal components (PCs) and showed a favorable clustering trend along PC1.

Conclusão

Important loadings for clustering were found in variables associated with carbohydrates, fatty acids, amino acids, water, caffeine, and chlorogenic acids. The DD-SIMCA model developed with two PCs achieved 100% sensitivity and specificity. The results showed that the developed method has the potential to innovate coffee safety control, representing an efficient new analytical approach for authentication.