Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1386254 | Carbohydrate Polymers | 2009 | 7 Pages |
A fast and non-destructive method to evaluate the condition of pulp and paper was developed. Partial least square regression (PLS-R) models based on near infrared (NIR) spectra and reference values for molecular weight, carbonyl group content and carboxyl group content were calculated for pulp hand sheets and rag papers.In this study, 110 pulp hand sheets were used and gave satisfactory models with high correlation coefficients (up to 0.97) during validation; whereas the test set validation (external validation) results were always better than those of cross-validation.Modeling of 267 historic rag paper samples was more demanding due to inherent variability of the material. Nevertheless, PLS-R models for the carbonyl group content, carboxyl group content and molecular weight with good correlation coefficients (up to 0.93) and low errors for cross-validation using average spectra of different paper samples were obtained. For carbonyl group content models with good correlation coefficient was also obtained without previous averaging. Joint models using both pulp hand sheets and rag papers were calculated for carboxyl and carbonyl group contents resulting in lower correlation coefficients then the single models.