Monitoramento da dessorção de água em polpas celulósicas por espectroscopia no NIR

• Ano de defesa: 2022
• Autor(a) principal: Medeiros, Dayane Targino de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Lavras; Programa de Pós-graduação em Ciência e Tecnologia da Madeira; UFLA; brasil; Departamento de Ciências Florestais
• Palavras-chave em Português: Aprendizado de máquinas; Monitoramento online; Controle de qualidade; Higroscopicidade; Sensoriamento remoto; Polpa celulósica - Umidade; Machine learning; Online monitoring; Quality control; Hygroscopicity; Remote sensoring; Cellulosic pulp - Moisture; Tecnologia e Utilização de Produtos Florestais
• Link de acesso: http://repositorio.ufla.br/jspui/handle/1/49518

The development of new solutions for improving product quality, optimizing the production process, and reducing costs is a fundamental factor in the search for competitiveness in companies in the pulp and paper sector. Near infrared spectroscopy (NIR) is a technique that has been presenting satisfactory performance and viability for technological evaluation in wood and cellulosic pulp due to its speed, precision, because it is a nondestructive analysis and that returns results in real time, thus enabling the rationalization of processes and consumption of raw material. In this context, the present study aimed to evaluate the potential of NIR spectroscopy associated with multivariate data analysis to estimate the variation of moisture in cellulosic pulp. Initially, the samples of four types of cellulosic pulp were saturated in different proportions by immersion in water. The proportion of fiber mass and water of the samples was determined before drying by the masses before and after saturation of the specimens. Then, the samples were submitted to air drying in a controlled environment and every 10% of the water mass loss was weighed on a precision analytical scale, followed by readings on the NIR spectrometer with Fourier transform equipped with an integration sphere. The weight and solid content of these materials were monitored during drying. The principal component analyses (PCA) and partial least squares regression (PLS-R) were applied to spectral signatures and reference data to estimate the humidity of the evaluated materials. The performance of the models was evaluated by cross-validation and independent batches. In the loss of mass, it was observed that the drying rate presented distinct behavior for each type of cellulosic pulp. And the NIR spectroscopy associated with PCA, was efficient in the differentiation of pulps and drying steps, in the same way as PLS-R by cross-validation and independent, and in the transfer of calibrations, was able to estimate the humidity of the materials under different conditions, with R²p ranging from 0.89 to 0.98 and, RMSEp between 5.16 and 18.35%, respectively. With this, the developed models have enough robustness to be used in the monitoring of the hygroscopic behavior of cellulosic pulp, subject to the variation of relative humidity, in real time during its transport, storage and commercialization.