Avaliação da capacidade preditiva de modelos quimiométricos baseados em espectroscopia NIR para a quantificação de terras raras em solos

• Ano de defesa: 2020
• Autor(a) principal: Pereira, Lavínia Hannah de Souza
• Orientador(a): Véras Neto, José Germano
• Banca de defesa: Simões, Simone da Silva, Barros, Sara Regina Ribeiro Carneiro de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual da Paraíba
• Programa de Pós-Graduação: Programa de Pós-Graduação em Química - PPGQ
• Departamento: Pró-Reitoria de Pós-Graduação e Pesquisa - PRPGP
• País: Brasil
• Palavras-chave em Português: Quimiometria; Regressão multivariada; Calibração multivariada
• Palavras-chave em Inglês: Multiway regression; Multiway calibration; Chemometrics
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::QUIMICA
• Link de acesso: http://tede.bc.uepb.edu.br/jspui/handle/tede/4051

The rare earth metals economic presence allows us to refer them as the "gold of 21st century”, relatively these elements are present in several types of soils, making possible an economic advance for regions that present an abundance of soils that contain these metals. So, technology advancement and production growth that mineral industry provides, the application of these chemical elements afford a technological evolution, due to their electronic, optical and also catalytic characteristics. Besides this, different methodologies have been applied to detect rare earths in soils, so that can offer low cost, don’t be destructive, and enable a high value added to these materials. The objective of this work was to measure chemometric models’ predictive capacity using Near Infrared Spectroscopy (NIR) technique to quantify rare earth in soils. The analysis was made in two kinds of soils (S and I) of three distinct regions at Borborema’s province, state of Pernambuco. These beginning tests was realized by ICP-OES technique, at UFRPE, the following analyses in Near Infrared – NIR at UEPB. The spectral data were obtained by NIR spectroscopy and previous treatments such as Multiplicative Scatter Correction (MSC), Standart Normal Variate (SNV) and Savitzky-Golay. To predict rara earth quantification was performed by Partial Least Squares regression (PLS) algorithm, and Successive Projection algorithms (SPA) -iPLS). Based on the results obtained, it is possible to make an analysis of the values predicted by PLS and SPA - iPLS, and verify that the figures of merit through the selection of variables with SPA-iPLS present significant accuracy values in comparison to the PLS model.