文献类型：期刊文献

中文题名：利用近红外光谱技术快速预测楸树木材抗弯性质

英文题名：Rapid Prediction of Bending Properties of Catalpa Bungei Wood by Near-Infrared Spectroscopy

作者：汪睿[1] 石兰兰[1] 王玉荣[1]

第一作者：汪睿

通信作者：Wang, YR[1]

机构：[1]中国林业科学研究院木材工业研究所,北京100091

年份：2023

卷号：43

期号：2

起止页码：557-562

中文期刊名：光谱学与光谱分析

外文期刊名：Spectroscopy and Spectral Analysis

收录：CSTPCD;;EI(收录号：20232314193708);Scopus;北大核心:【北大核心2020】；CSCD:【CSCD2023_2024】；PubMed;

基金：国家“十三五”重点研发计划课题(2017YFD0600201);中央级公益性科研院所基金项目(CAFYBB2018GD001)资助。

语种：中文

中文关键词：楸木;近红外光谱;抗弯性质;木材切面;预处理方法;采谱点数

外文关键词：Catalpa bungei wood;Near-infrared spectroscopy;Bending property;Wood sections;Pretreatment;Sampling point

分类号：S781

摘要：楸树(Catalpa bungei)木材纹理通直、材性优良、用途广泛,是中国特有的珍贵材树种。研究木材重要的力学性质——抗弯性质的快速测定方法可以为楸树木材的遗传改良及加工利用提供科学依据。以楸树无性系新品种“洛楸1号”、“洛楸4号”和“天楸2号”为试验材料,依据国家标准抗弯性质测试方法,测定楸树木材抗弯强度(MOR)和抗弯弹性模量(MOE)。利用近红外光谱(NIRs)分析结合偏最小二乘法(PLS)对新选育的三个楸树无性系的抗弯性质进行预测,探究基于不同采谱切面、不同预处理方法以及不同采谱点的最佳建模方法。研究结果表明,基于两切面平均光谱建立的抗弯强度预测模型的相关系数和相对分析误差最高为0.843和1.88,建立的抗弯弹性模量预测模型的相关系数和相对分析误差最高为0.846和1.88。选用两切面光谱,预处理方法按抗弯强度模型性能排序为多元散射校正与卷积平滑结合算法(MSC+S-G)>二阶导数与卷积平滑结合算法(2^(nd)Der+S-G)>一阶导数与卷积平滑结合算法(1^(st)Der+S-G),预处理方法按抗弯弹性模量模型性能排序为MSC+S-G>1^(st)Der+S-G>2^(nd)Der+S-G。使用一点采谱法建立的抗弯强度和抗弯弹性模量预测模型相关系数比五点采谱法的分别降低5.93%和2.96%。综上所述,近红外光谱可以用于预测珍贵材楸木的抗弯强度和抗弯弹性模量。采用不同切面、预处理方法和采谱点数建立的模型,建模结果有一定差异。得出了楸树木材抗弯强度和抗弯弹性模量的最佳建模方法。基于径切面和弦切面平均光谱建立的抗弯强度和抗弯弹性模量近红外模型效果最佳。MSC+S-G是最适用于楸树木材抗弯性质的预处理方法。五点采谱法模型精度较高,但对大量样品抗弯性质快速估算时,可以降低采谱点数量,仅采集中间部位即载荷加载部位一个光谱点以减少采谱工作量,提高楸树木材抗弯性质快速评估效率。
Catalpa bungei has the advantages of straight texture,excellent material,and versatility characteristics and it is a precious wood species unique to China.Bending property,an important mechanical property of wood,research on its rapid determination method can provide a scientific basis for genetic improvement,processing,and utilization of Catalpa wood.The“Luoqiu 1”,“Luoqiu 4”and“Tianqiu 2”of the new C.bungei clones were used as the experiment materials.The modulus of rupture(MOR)and modulus of elasticity(MOE)was determined according to the national standard bending property test method.Near-infrared spectroscopy(NIRs)combined with the partial least squares(PLS)method was used to predict the bending properties of three newly bred C.bungei clones.The best modeling method based on different wood sections,pretreatment methods,and the number of sampling points were explored.The results indicated that the maximum Rpand RDP of the MOR model based on the average spectra of two sections were 0.843 and 1.88,and the maximum Rpand RDP of the MOE prediction model were 0.846 and 1.88.In descending order of accuracy of MOR models based on average sections,pretreatments were:MSC+S-G,2^(nd)Der+S-G,and 1^(st)Der+S-G.In descending order of accuracy of MOE models based on average sections,pretreatments were:MSC+S-G,1^(st)Der+S-G,and 2^(nd)Der+S-G.In conclusion,NIRs can be used to predict the MOR and MOE of valuable C.bungei wood.Models established with different sections,pretreatments,and the number of sampling points have certain differences in modeling results.This paper obtained the best modeling methods for the MOR and MOE of C.bungei wood.NIR models of MOR and MOE based on average spectra of radial and tangential sections were the best.MSC+S-G was the most suitable pretreatment method for the bending properties of C.bungei wood.The five-point sampling method has the highest model accuracy.The number of sampling points can be reduced to quickly estimate the bending property of a large number of samples.It is possible to collect only one spectra point in the middle part,the loading position,to reduce the workload of collecting spectra and improve the efficiency of rapid evaluation of the bending property of C.bungei wood.