Simultaneous modelling of process variables and raw material properties as measured by NIR. A case study from cellulose production

Summary

The main goal of the present paper was to investigate the potential of near infrared spectroscopy (NIR) to be used in modelling of pulp properties in the paper industry. An experimental design based on a split plot structure was used for generating the data. The factors considered were cooking recipe, cooking time and chips quality and the response was Kappa No. in sulphite pulp. NIR spectra were measured on the chip samples, and transformed to principal components. The scores from these components were used in an ANOVA model together with the other design variables. The first step in the modelling work was to establish a benchmark model, which included only chips category, and not the NIR spectra themselves. Then the scores from the principal component analysis were included in the model. One principal component was found to be significant for the prediction of Kappa No. Prior to the model building process, a thorough investigation of the principal component analysis was performed, including a discriminant analysis of the scores. The main conclusions from this work are that it is possible to categorize chips according to scores on corresponding NIR spectra, and to replace chips category with these scores in ANOVA models for Kappa No. in sulphite pulp. (c) 2005 Published by Elsevier B.V The main goal of the present paper was to investigate the potential of near infrared spectroscopy (NIR) to be used in modelling of pulp properties in the paper industry. An experimental design based on a split plot structure was used for generating the data. The factors considered were cooking recipe, cooking time and chips quality and the response was Kappa No. in sulphite pulp. NIR spectra were measured on the chip samples, and transformed to principal components. The scores from these components were used in an ANOVA model together with the other design variables. The first step in the modelling work was to establish a benchmark model, which included only chips category, and not the NIR spectra themselves. Then the scores from the principal component analysis were included in the model. One principal component was found to be significant for the prediction of Kappa No. Prior to the model building process, a thorough investigation of the principal component analysis was performed, including a discriminant analysis of the scores. The main conclusions from this work are that it is possible to categorize chips according to scores on corresponding NIR spectra, and to replace chips category with these scores in ANOVA models for Kappa No. in sulphite pulp. (c) 2005 Published by Elsevier B.V