Study On The Determin Of Natural Cellulose Cleaning Pulps By Near Infrared Reflectance Spectroscopy

In this letter, near-infrared diffuse reflectance spectroscopy is used to determine the properties of natural cellulose cleaning pulps (mainly cotton and wood). Using different types of NIR spectroscopy to determine their degree of polymerization (DP), a-cellulose content and pentosan content after different pretreatment. Establish a qualitative classification model of mixed samples and inspect its model classification effect, as well as their quantitative prediction models and investigate the prediction accuracy of the models, correct the online feasibility of NIR spectroscopy. It provides a reliable reference for the on line testing of properties in pulp production process using near infrared spectroscopy technology. The specific conclusions are as follows:1, DP data of the samples are determined by standard method GB/T9107-1999and the spectrums of corresponding samples are collected by Sup-2700NIR spectroscopy. The qualitative classification models of mixed samples are established using the principal component analysis (PCR) and cluster class independent soft mode (SIMCA) method, it’s recognition rate for cotton, wood pulp samples can reach100%. The quantitative models of cotton, wood pulp samples alone and mixed samples are established using partial least squares (PLS) method, the corresponding RC and RP are:0.991,0.983;0.886,0.877, and RMSEP143and51respectively. The prediction precision and accuracy of individual samples are closed to national standard method requirement.2, The spectrum of corresponding samples are collected by FT-NIR Analyzer Fourier Transform NIR spectroscopy before an tablet pretreatment. The quantitative models of cotton, wood pulp samples respectively and mixed samples are established using partial least squares (PLS) method, the corresponding RC, RP and RMSEP are:0.950,0.924,419;0.845,0.839,137;0.944,0.873,316o Despite its good correlation coefficient, the prediction precision is much bigger which can’t meet national standard method requirement. So the models cannot be used to predict DP of the samples accurately, neither the spectrum could be collected in a still way.3, The spectrum of corresponding samples are collected by JDSU NIR spectroscopy before an tablet pretreatment in a still and rotate state respectively, the rotating style is much better than the stationary style. the model of wood sample in a rotate state has an excellent correlation and prediction with an RC, RP and RMSEP of:0.869,0.898,53.170. The model is feasible for determining DP of natural cellulose in clean pulping process. Establish two quantitative models of cotton pulp samples according to their level of DP, though poor in correlation, the models’prediction accuracy are significantly increased. 4, a-cellulose content of the samples is determined by standard method GB/T9107-1999and the spectrums of corresponding samples are collected by Sup-2700NIR spectroscopy. The quantitative models of cotton, wood pulp samples alone and mixed samples are established using partial least squares (PLS) method, the corresponding RC,RP and RMSEP are:0.928,0.939,0.011;0.885,0.855,0.020;0.955,0.954,0.020. The classification samples are much better than that of mixed samples, and they are feasible for determininga-cellulose content of natural cellulose in clean pulping process.5, Pentason content of the wood samples are determined by standard method GB/T9107-1999and the spectrums of corresponding samples are collected by Sup-2700NIR spectroscopy. The quantitative models of cotton, wood pulp samples alone and mixed samples are established using partial least squares (PLS) method, the corresponding RC,RP and RMSEP are:0.929,0.968,0.007. The prediction accuracy of the model is closed to the laboratory measurement results. It shows that using near infrared spectroscopy method to detect the pentosan content is feasible.