Estimating Wood Speices And Chemical Compositions In Pulping By Near Infrared Spectroscopy

Wood chip as the main raw material for papermaking, its properties influence the product quality directly. We can filter wood chip according its properties and monitor online, in order to control and predict the pulp quality in time. We can control the parameters and dosing volume refining the process to achieve the best wood utilization and yield. The near infrared spectroscopy as a rapid non-destructive testing methods and has rich wood absorption spectra, we can predict wood properties quickly through the establishment of near-infrared spectroscopy model to achieve its online analysis.In the paper, we use the E.urophylla×E.grondis, Eucalyptus urophylla LH1, Eucalyptus urophylla U6, Eucalyptus globules, Eucalyptus guangxi, Acacia mangium and Acaciacrassicarpa as the raw material.Firstly, near-infrared spectra of wood flour and chip were collected by Anatris and Hadamard near-infrared spectrometer. Then the wood properties containing holocellulose, klason lignin and benzene-alcohol extractive were tested and the NIR of holocellulose and klason lignin were collected by Anatris near-infrared spectrometer. Each timber of the chemical component content and absorbance spectra were different by comparative study.Finally, the absorbance spectra band of holocellulose, klason lignin and water was 4730~4798 cm-1, 6720~6765 cm-1 and 4177~4285 cm-1, 5792~5900 cm-1, 6861~6911 cm-1 and 4700~5200 cm-1 screened out by observation. The absorbance spectra point of holocellulose and klason lignin was 4740?4756?4767?4787?6719?6734?6765 cm-1;4289?4304?4382?4405?5793?5827?5863?5920?6850?6885?6908?6923cm-1 screened out by Competitive Adaptive Reweighted Sampling.Firstly, near-infrared spectra of 86 samples were collected. Then partial least squares discriminant analysis(PLS-DA) was used to establish a classification model of eucalyptus and acacia. The results showed that the regression lines between predicted value and actual value of calibration set and validation set were in good agreement, and the coefficient of determination(R2) were 0.99 and 0.97, which showed the model is better. What's more, the recognition of unknown samples was 100 %. MSC and Savitzky-Golay smoothing method were used to pretreat the spectral of 7500~4000 cm-1, and discrimination model was established based on the principal component analysis. Results showed that the calibration and validation model was perfect. MSC and Savitzky-Golay smoothing method were used to pretreat the spectral of 5500~6500 cm-1, and discrimination model was established based on the principal component analysis. Results showed that the recognition rate of the calibration and validation model was 100 %. Moreover, the chemometrics algorithm based on near-infrared spectra could indentify tree species quickly. Therefore, Hadamard near-infrared spectrometer can also be used to identify the eucalyptus and acacia.The near-infrared model of cellulose, lignin and benzene-alcohol extractive were established by 44 species of Eucalyptus and Acacia wood chip and flour. The near-infrared wood flour model of cellulose, lignin and benzene-alcohol extractive were established through TQ Analyst 8.0 partial least sqaure method, then the model was verified by cross valiation method, the value of RCV were 0.9812, 0.9226, 0.9709, RMSECV were 0.1890, 0.6019, 0.2288, RPD were 5.1815, 2.5923, 4.1756. The model was verified by predicting other samples without the modeling, and the deviation between predict value and real value was small. Finally, the predicted result was further confirmed by paired sample T-test, when the significance level was set to 0.05, no significant difference was found between them. The near-infrared wood chip model of cellulose, lignin and benzene-alcohol extractive were established through TQ Analyst 8.0 partial least sqaure method, then the model was verified by cross valiation method, the value of RCV were 0.9046, 0.9052, 0.9305, RMSECV were 0.5173, 0.6601, 0.3425, RPD were 2.3460, 2.3530, 2.7300. Compared the model results of the Anatris and Hadamard near-infrared spectrometer, the former was better, and the latter model was within the acceptable range.