基于LASSO法的桉木-相思混合制浆原料近红外分析模型的建立  

Establishment of Near Infrared Analysis Models of Eucalyptus-Acacia Mixed Pulpwood Materials Based on LASSO Algorithm

作　　者：吴珽 梁龙[1] 朱北平[1] 邓拥军[1] 房桂干[1] WU Ting;LIANG Long;ZHU Beiping;DENG Yongjun;FANG Guigan(Institute of Chemical Industry of Forest Products,CAF,National Engineering Lab.for Biomass Chemical Utilization,Key Lab.of Chemical Engineering of Forest Products,National Forestry and Grassland Administration,Key Lab.of Biomass Energy and Material,Jiangsu Province,Jiangsu Co-Innovation Center of Efficient Processing and Utilization of Forest Resources,Nanjing 210042,China;Gold East Paper(Jiangsu)Co.,Ltd.,Zhenjiang 212132,China)

机构地区：[1]中国林业科学研究院林产化学工业研究所,生物质化学利用国家工程实验室,国家林业和草原局林产化学工程重点实验室,江苏省生物质能源与材料重点实验室,江苏省林业资源高效加工利用协同创新中心,江苏南京210042 [2]金东纸业(江苏)股份有限公司,江苏镇江212132

出　　处：《林产化学与工业》2020年第5期83-90,共8页Chemistry and Industry of Forest Products

基　　金：中国博士后科学基金资助项目(2019M661780);国家重点研发计划资助项目(2017YFD0601005)。

摘　　要：为提高国内木浆产量和利用率,缓解制浆造纸行业优质原料紧缺的现状,针对我国南方特定的制浆原料模式——桉木-相思混合原料进行快速分析研究。采集了175个桉木-相思混合原料和45个单一材种原料的近红外光谱,明确其混合程度和化学成分含量。通过平滑、矢量归一化(V-Norm)、多元散射校正(MSC)、一阶导数(1st Der)和二阶导数(2nd Der)等方法组合预处理原始光谱,结合LASSO算法,确定了混合程度(桉木质量分数)、综纤维素、Klason木质素、聚戊糖、苯醇抽出物和1%NaOH抽出物含量分析模型的建模方法,并建立了相应的模型。建模过程中最优调整参数(μ)分别为13.62、18.30、6.39、9.64、7.49和12.07。6个模型的预测均方根误差(RMSEP)值分别为1.93%、0.61%、0.51%、0.80%、0.28%和0.41%。绝对偏差范围(AD)分别为-3.19%~3.24%、-0.96%~1.01%、-0.89%~0.84%、-1.37%~1.46%、-0.43%~0.39%和-0.58%~0.60%。模型适应性好,能够满足制浆造纸工业需求,同时也证实了LASSO法用于混合制浆原料分析的可行性。In order to improve the domestic pulp output and utilization rate,and alleviate the shortage of high-quality materials in the pulping and paper-making industry,a rapid analysis research was conducted on the specific material model of pulp making—Eucalyptus-Acacia mixed with pulpwood in southern China.The near-infrared spectra of 175 Eucalyptus-Acacia mixed samples and 45 single-species samples were collected.The mixing degree and chemical composition content of all samples were analyzed.The original spectra were preprocessed by the combined pretreatment methods of smoothing,vector normalization(V-Norm),multiple scattering correction(MSC),first derivative(1st Der)and second derivative(2nd Der).Combined with the least absolute shrinkage and selection operator(LASSO)algorithm,the mixing degree,holocellulose,Klason lignin,pentosan,benzene-alcohol extractives and 1%NaOH extractives content models were built.The optimal adjustment parameters determined during the modeling process were 13.62,18.30,6.39,9.64,7.49,and 12.07.The RMSEP values of the six models were 1.93%,0.61%,0.51%,0.80%,0.28%,and 0.41%,respectively.The absolute deviation ranges were-3.19%-3.24%,-0.96%-1.01%,-0.89%-0.84%,-1.37%-1.46%,-0.43%-0.39%,-0.58%-0.60%.The models have good adaptability and could meet the needs of the pulping and paper-making industry.It also confirmed the feasibility of the LASSO algorithm for the analysis of mixed pulpwood materials.

关 键 词：近红外光谱 制浆造纸 LASSO算法 预处理 桉木-相思 

分 类 号：TQ35[化学工程] TS721[轻工技术与工程—制浆造纸工程]