Prediction of strength parameters for softwood kraft pulps. Multivariate data analysis based on orthogonal signal correction and near infrared spectroscopy

The present work presents a near-infrared (NIR) study undertaken to explore the relationship between the choice of softwood raw material and the prope rties of the resulting kraft pulps expressed in terms of physical parameter s for the pulps and strength properties for the corresponding handsheets. T he kraft pulps were made from 20 different types of wood samples, which wer e chosen according to an experimental design. NIR spectra were recorded for the wood chips and the fully bleached pulp samples. To evaluate the relati on between NIR and the end properties, multivariate data analysis was used to generate prediction models for fiber properties and strength parameters. A new method was introduced where the information in the NIR matrix which i s orthogonal to the matrix described by the individual properties was elimi nated. The results clearly demonstrated the power of this "soft" target rot ation technique and the power of using two-block multivariate techniques to predict pulp and paper properties from NIR data. Using this new method, ca lled orthogonal signal correction (OSC), it was found that the NIR spectra of milled wood chips have nearly the same predictive ability as those of th e bleached pulps. The PLS models for strength parameters based on NIR spect ra of pulp samples used between 95% and 98% of the variation of the pretrea ted NIR spectra to explain between 75 and 92% of the variation of the stren gth parameters using one PLS component. The predictive ability was good, co rresponding to Q(2)(cum) values ranging from 73.3% to 90.8%. The PLS models based on OSC treated NIR spectra of wood samples used between 77 and 90% o f the variation of the OSC-corrected NIR spectra to explain between 71 and 88% of the variation of the strength parameters using one PLS component. Th e predictive ability corresponded to Q(2)(cum) values ranged from 62.1% to 87.2%.