| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.38 MB | Adobe PDF |
Advisor(s)
Abstract(s)
As propriedades das matérias primas utilizadas na produção de painéis de aglomerado de
madeira, nomeadamente do material lenhocelulósico, dos adesivos utilizados, bem como dos
aditivos, são fatores determinantes da qualidade dos mesmos. Desta forma o valor de pH e
capacidade tampão da madeira utilizada são parâmetros relevantes no sentido em que irão
influenciar a cura da resina adicionada à mesma durante a etapa de prensagem do painel. A
espetroscopia de infravermelho próximo (NIR) caracteriza-se como uma técnica de interesse
no sentido em que a sua implementação na linha de produção irá acelerar e facilitar a
otimização do processo produtivo.
Assim, o objetivo do presente trabalho passou, numa fase inicial, pelo aprimoramento do
método de referência para medição das propriedades em estudo – valor de pH e capacidade
tampão – já implementado na empresa, bem como pela construção de modelos preditivos com
recurso a métodos quimiométricos, nomeadamente a regressão por mínimos quadrados parciais
(PLS). Os espetros foram adquiridos tanto em modo de transmitância (soluções líquidas), como
foram também analisadas amostras de partícula de madeira (sólidas) em modo de refletância.
Primeiramente, foram estudados diversos parâmetros do método de referência, sendo que
foram testados três procedimentos para a extração dos compostos da partícula de madeira.
Assim, chegou-se a um método final que consiste na adição de 300 g de água destilada à
madeira, que é ebulida com um tempo de refluxo de 30 min, sendo que foi acrescentado um
passo de lavagem da partícula. A água de lavagem é então utilizada para a diluição da solução
líquida filtrada e realizadas as medições de pH e capacidade tampão.
Tendo sido otimizado o método de extração foram realizados ensaios à partícula sem
modificações, a frações peneiradas da partícula, bem como com adição de buffers/captadores,
e ainda à partícula com diferentes valores de humidade absoluta.
Procedeu-se à construção dos modelos para a determinação do valor de pH e capacidade
tampão ácida (CTA) e básica (CTB), relacionando os espetros adquiridos tanto das soluções
líquidas extraídas da amostra (transmitância), como dos adquiridos a partir das amostras
sólidas (refletância) com os valores obtidos com o método de referência. A qualidade dos
mesmos foi avaliada com base nos valores da raiz quadrada do erro médio quadrático de
validação cruzada (RMSECV) e do coeficiente de determinação (R2), que serão idealmente
próximos de zero e acima de 0,95, respetivamente. Construíram-se, então, três modelos para cada modo de aquisição de espetros - de pH, CTA e
CTB -, sendo que os resultados mais promissores foram os obtidos para o modo de
transmitância, em particular o de previsão de pH com um RMSECV de 0,0917 e um R2 de
0,9022. Para o modo de refletância também o modelo de previsão de pH foi o que melhores
resultados apresentou, com a obtenção de um RMSECV de 0,12 e de um R2 de 0,7139.
Conclui-se que o modo de refletância apresenta resultados menos satisfatórios
comparativamente à transmitância, resultado este já esperado, tendo em conta que os efeitos
de dispersão da luz são mais acentuados e o facto de esta aquisição ser feita em movimento,
não permitindo uma distância constante entre a sonda e a amostra.
The properties of the raw materials used in the production of wood particleboard, namely the lignocellulosic material, the adhesives used, as well as the additives, are determining factors in their quality. Thus, the pH value and buffering capacity of the wood used are relevant parameters as they will influence the curing of the resin added to it during the panel pressing stage. Near-infrared spectroscopy (NIR) is characterized as a technique of interest in that its implementation on the production line will accelerate and facilitate the optimization of the production process. Thus, the objective of the present work initially involved improving the analytical measurement method for the properties under study – pH value and buffering capacity – already implemented in the company, as well as constructing predictive models using chemometric methods, namely Partial Least Squares Regression (PLS). The spectra were acquired both in transmittance mode (liquid samples) and wood particle samples (solid) were also analyzed in reflectance mode. First, several parameters of the reference method were studied, with three procedures for the extraction of compounds from the wood particle tested. Thus, a final method was reached which consists of adding 300 g of distilled water to the wood, which is boiled with a reflux time of 30 minutes, with a washing step of the particle added. The washing water is then used to dilute the filtered liquid solution, and pH and buffering capacity measurements are taken. With the extraction method optimized, tests were performed on the unmodified particle, sieved fractions of the particle, as well as with the addition of buffers/scavengers, and also on the particle with different absolute moisture values. Models were then constructed for the determination of the pH value, acidic buffering capacity (CTA) and basic buffering capacity (CTB), relating the spectra acquired both from the liquid solutions extracted from the sample (transmittance) and from the solid samples (reflectance) with the values obtained by the reference method. Their quality was evaluated based on the values of the Root Mean Square Error of Cross-Validation (RMSECV) and the coefficient of determination (R²), which should ideally be close to zero and above 0.95, respectively. Thus, three models were constructed for each spectrum acquisition mode - pH, CTA, and CTB - with the most promising results obtained for the transmittance mode, particularly the pH prediction with an RMSECV of 0.0917 and an R² of 0.9022. For the reflectance mode, the pH prediction model also presented the best results, with an RMSECV of 0.12 and an R² of 0.7139. It is concluded that the reflectance mode presents less satisfactory results compared to transmittance, a result that was already expected, considering that the light scattering effects are more pronounced and the fact that this acquisition is done in motion, not allowing a constant distance between the probe and the sample.
The properties of the raw materials used in the production of wood particleboard, namely the lignocellulosic material, the adhesives used, as well as the additives, are determining factors in their quality. Thus, the pH value and buffering capacity of the wood used are relevant parameters as they will influence the curing of the resin added to it during the panel pressing stage. Near-infrared spectroscopy (NIR) is characterized as a technique of interest in that its implementation on the production line will accelerate and facilitate the optimization of the production process. Thus, the objective of the present work initially involved improving the analytical measurement method for the properties under study – pH value and buffering capacity – already implemented in the company, as well as constructing predictive models using chemometric methods, namely Partial Least Squares Regression (PLS). The spectra were acquired both in transmittance mode (liquid samples) and wood particle samples (solid) were also analyzed in reflectance mode. First, several parameters of the reference method were studied, with three procedures for the extraction of compounds from the wood particle tested. Thus, a final method was reached which consists of adding 300 g of distilled water to the wood, which is boiled with a reflux time of 30 minutes, with a washing step of the particle added. The washing water is then used to dilute the filtered liquid solution, and pH and buffering capacity measurements are taken. With the extraction method optimized, tests were performed on the unmodified particle, sieved fractions of the particle, as well as with the addition of buffers/scavengers, and also on the particle with different absolute moisture values. Models were then constructed for the determination of the pH value, acidic buffering capacity (CTA) and basic buffering capacity (CTB), relating the spectra acquired both from the liquid solutions extracted from the sample (transmittance) and from the solid samples (reflectance) with the values obtained by the reference method. Their quality was evaluated based on the values of the Root Mean Square Error of Cross-Validation (RMSECV) and the coefficient of determination (R²), which should ideally be close to zero and above 0.95, respectively. Thus, three models were constructed for each spectrum acquisition mode - pH, CTA, and CTB - with the most promising results obtained for the transmittance mode, particularly the pH prediction with an RMSECV of 0.0917 and an R² of 0.9022. For the reflectance mode, the pH prediction model also presented the best results, with an RMSECV of 0.12 and an R² of 0.7139. It is concluded that the reflectance mode presents less satisfactory results compared to transmittance, a result that was already expected, considering that the light scattering effects are more pronounced and the fact that this acquisition is done in motion, not allowing a constant distance between the probe and the sample.
Description
Financiado
Keywords
Near infrared spectroscopy pH Buffer capacity PLS Espetroscopia de infravermelho próximo Capacidade tampão