Premium
Multivariate calibration models using NIR spectroscopy on pulp and paper industrial applications
Author(s) -
Antti Henrik,
Sjöström Michael,
Wallbäcks Lars
Publication year - 1996
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/(sici)1099-128x(199609)10:5/6<591::aid-cem474>3.0.co;2-l
Subject(s) - partial least squares regression , multivariate statistics , near infrared spectroscopy , pulp (tooth) , chemometrics , calibration , spectroscopy , test set , environmental science , mathematics , pulp and paper industry , biological system , computer science , statistics , engineering , machine learning , optics , medicine , physics , pathology , quantum mechanics , biology
A goal for the pulp and paper industry is to get a fast and reliable characterization of raw materials as wood and pulp compositions. One possibility for this is near infrared reflectance (NIR) spectroscopy combined with multivariate analysis. In the first part of this study the possibility to make predictions of mixtures of wood chips from three different wood species (Swedish pine, Swedish spruce and Polish pine) is investigated based on NIR spectroscopy. Mixture design and Partial least squares projections to latent structures (PLS) were used for the multivariate calibration modeling. The calculated model was validated both internally and with an external test set. The result was a PLS model with a Q 2 = 0.91 according to cross‐validation and good prediction of the test set objects, Q 2 test set = 0·78. In the second part NIR spectroscopy is used to characterize a series of pulp samples. The pulps were also characterized by seventeen traditionally measured pulp properties. PLS was used for the calibration model and internal as well as external validation were done. The resulting PLS model for the 17 pulp properties gave an overall Q 2 = 0·61 according to cross‐validation. Predictions of the test set objects show that most of the properties are well described by the model. © 1996 by John Wiley & Sons, Ltd.