Use of NIR spectroscopy and multivariate regression for prediction of pentosan content in wood pulp

Wood is a complex material whose main chemical constituents are cellulose, hemicellulose, and lignin. These components are studied in various industries after the wood has been processed by chemical or mechanical methods. For the paper industry, it is relevant to determine the pentosan content in cellulose pulp because it indicates the degree of retention of hemicellulose. Hemicellulose contributes to the resistance, increasing the yield of the pulp, and therefore, high pentosan content is desirable. In this way, this research focused on the determination of the pentosan content in hard and softwood pulps between 0.81 to 18.4%. The pentosan content can be directly determined by a chemical method, although these conventional methods are long, expensive, generate a high amount of corrosive waste, and are not recommended for routine analysis. Therefore, in this research, an alternative method was developed using near-infrared spectroscopy together with partial least squares regression to predict the pentosan content in pulps. This new method is fast, inexpensive, analysis is direct and non-destructive. Finally, the pentosan calibration model was validated by cross-validation and the predicted external samples were quantified with precision between 0.008 and 0.043 and accuracy between 3.9 and 12.2%, while SEP has a variability of 1.267% of pentosan for this model.