Aplicação da espectroscopia no infravermelho próximo para avaliação do carvão vegetal

• Autor(a) principal: Costa, Lívia Ribeiro
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/12992
• topic: Carvão vegetal; Espectroscopia no NIR; Regressão PLS; Principal Component Analysis (PCA); Partial Least Squares Discriminant Analysis (PLS-DA); Charcoal; NIR spectroscopy; Regression PLS; Recursos Florestais e Engenharia Florestal

For controling the quality of their products, charcoal steels mills require technologies capable of predicting the characteristics of their bioreducing agent quickly and reliably, since conventional analyzes are expensive and time-consuming. Near infrared spectroscopy (NIR) has been successfully applied to determine charcoal properties, allowing material classification in real time. There are no studies in the literature that applied the technique to evaluate the carbonization temperature and the density of the carbonaceous material. Thus, this study aimed to establish multivariate models to estimate gravimetric carbonization yield (GCY), apparent relative density (ARD) and final carbonization temperature (FTC of Eucalyptus charcoal by NIR spectroscopy. Eucalyptus clones from commercial plantations managed for energy purposes and pulp and paper production were used. Wood prismatic specimens with nominal dimensions of 25 mm x 25 mm x 80 mm were carbonized at final temperatures of 400°C, 500°C, 600°C and 700°C. NIR spectra measured directly on 160 charcoal specimens were correlated with GCY and ARD values obtained through conventional laboratory analyzes. Principal component analysis (PCA), partial least squares regression (PLS-R) and partial least squares discriminant analysis (PLS-DA) were utilized based in spectral and experimental information. The NIRS technique associated with PLS-R was able to predict FTP and GCY presenting cross-validation coefficients (R²cv) of 0.96 and 0.85, respectively. It was not possible to predict apparent relative density based on charcoal spectral signature. Mathematical treatments did not provide better calibrations for the charcoal properties evaluated. Crossed and independent validations presented similar statistics, confirming NIR spectroscopy capacity associated with multivariate statistics at controlling charcoal quality. Specimen’s classifications into carbonization temperature groups through PLS-DA obtained 100% correct classification, except for 500°C temperature (97.5%). From the developed models, it is possible to predict charcoal characteristics such as gravimetric carbonization yield and final pyrolysis temperature. This work will serve as a reference for the development of new studies and applications of the technique NIR in charcoal.