Identificação de microplásticos no ambiente e detecção de impurezas em café por imageamento hiperespectral na região do infravermelho próximo (HSI-NIR) e quimiometria

Near-infrared hyperspectral imaging (HSI-NIR) and chemometric classification models were applied for the detection of impurities or contaminants at the micro and millimeter level with minimal sample preparation, as an alternative to the more laborious and subjective methods commonly used, in two topics of societal and technical-scientific relevance.The first topic is related to the global concern of microplastic (MP) pollution, and there is the demand for the development of analytical methods to detect them.The use of HSI-NIR combined with Soft Independent Modelling of Class Analogy (SIMCA) is described to automatically identify MP of the classes polyethylene (PE), polypropylene (PP), polyamide-6 (PA), polyethylene terephthalate (PET) and polystyrene (PS) in beach sand with minimum sample preparation.An in-house comprehensive spectral dataset, including from pure to weathered MP, was used to build the classification models.SIMCA sensitivity and specificity were over 99 % for all classes.Models were validated with hundreds of primary and secondary MP collected in the environment.The effect of particle size, color and weathering are discussed.The method was applied to environmental samples, identifying thousands of particles without prior visual sorting, which is a time-efficiency method contribution.The second topic describes the proposal for a fast method for the determination of the maximum allowed 1% w/w of impurities in roasted and ground coffee, through sample direct analysis by HSI-NIR.Four classification models using Partial Least Squares Discriminant Analysis (PLS-DA) were developed for coffee and impurities classes, obtaining from 97 to 99% specificity and sensitivity for all models, despite the spectral similarity between species, noisy spectral region and within the technique instrumental limits.The classification prediction results were used for quantification by the ratio between the image pixels classified as impurities and the total image pixels (Percentage of Impurity Pixels, PPI).The effect of edges at the interfaces between species in predicting and counting pixels of different classes was demonstrated.Although impurities have been identified at the chemical maps, the PPI values were not repetitive for different samples with the same impurity mass content, and coffee granulometry was the main cause.Limitations regarding the use of images in direct quantitative analysis of segregated solids were demonstrated. SUMÁRIO