1. Institute of Chemical Industry of Forest Products, Chinese Academy of Forestry, National Engineering Lab for Biomass Chemical Utilization; Key Lab of Biomass Energy and Material, Jiangsu Province, Nanjing 210042, China
2. Collaborative Innovation Center for High Efficient Processing and Utilization of Forestry Resources, Nanjing Forestry University, Nanjing 210037, China
3. College of Light Industry Science and Engineering, Nanjing Forestry University, Nanjing 210037, China
Abstract:In recent years, with the advance of forest and paper integration strategy, we often use mixed raw materials pulping. It is difficult to realize the rapid analysis of mixing degree and chemical composition content of raw materials, which has become the bottleneck constraints of pulping industry development. In order to solve this problem, the research chose the widely used poplar-eucalyptus wood mixed raw materials as study object, the near infrared spectrums of 131 poplar-eucalyptus wood samples which poplar content was artificially controlled and 30 single poplar and eucalyptus wood samples were collected with Fourier near infrared spectrometer, then the content of holocellulose, pentosan and Klason lignin was measured by chemical methods. The near-infrared spectra of these major chemical components are concentrated in the 7 600~4 000 cm-1 interval. The model of poplar content and the model of pentosan content were established by LASSO(the least absolute shrinkage and selection operator) algorithm combined with spectral data of 7 600~4 000 cm-1 which was pretreated by smoothing, standard normal variate and first derivative. The holocellulose content model was established with LASSO algorithm combined with same range of spectral data which was pretreated by smoothing, standard normal variate and second derivative. The Klason lignin content model was developed with the same algorithm , the same range of spectral data with the pretreatment of smoothing, multipicative scatter correction and second derivative. Poplar content, holocellulose, pentosan and Klason lignin models have root mean square error of prediction of 1.82%, 0.52%, 0.67% and 0.59% respectively. Absolute deviation (AD) range were -3.01%~2.94%, -0.91%~0.83%, -0.91%~1.07%, -0.79%~0.92%. The models have good performance better than the traditional partial least squares models that can be applied in actual industrial production.
Key words:Near-infrared spectroscopy technology; LASSO algorithm; Pretreatment; Mixed raw materials
[1] TIAN Chao, XUE Chong-yun, CHEN Bin(田 超, 薛崇昀, 陈 彬). China Pulp & Paper(中国造纸), 2010, 29(12): 10.
[2] Muhammad A J, Ong S S, Ratnam W. Journal of Forestry Research, 2017, (2): 1.
[3] Tsuchikawa S, Schwanninger M. Applied Spectroscopy Reviews, 2013, 48(7): 560.
[4] Stirling R. The Forestry Chronicle, 2013, 89(5): 654.
[5] ZHONG Yong-cui, YANG Li-wei, QIU Yun-qi, et al(钟永翠, 杨立伟, 邱蕴绮, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2017, 37(6): 1771.
[6] Balabin R M, Safieva R Z, Lomakina E I. Microchemical Journal, 2011, 98(1): 121.
[7] Bouslamti M A, Irle M A, Belloncle C, et al. International Wood Products Journal, 2013, 4(4): 116.
[8] Fagan C C, Everard C D, Mcdonnell K. Bioresource Technology, 2011, 102(8): 5200.
[9] Mora C R, Schimleck L R. Wood Science and Technology, 2010, 44(4): 561.
[10] JIANG Cui-xia, LIU Yu-ye, XU Qi-fa(蒋翠侠, 刘玉叶, 许启发). Journal of Management Sciences in China(管理科学学报), 2016, 19(3): 107.
[11] Tibshirani R. Journal of the Royal Statistical Society, 2011, 73(3): 267.
[12] He Wenming, Hu Huiren. Bioresource Technology, 2013, 140(7): 299.