中文摘要：

相思树是一种速生纸浆材,苯醇抽提物含量对木材的制浆得率有一定影响。近红外光谱分析技术能对木材的化学成分含量进行低成本快速检测。多模型方法是一种预测效果好且易于掌握的近红外光谱分析建模方法,已被用于建立相思树、毛白杨和欧美杨某些化学成分含量的近红外光谱预测模型,取得较好的建模效果。先用多模型方法建立了相思树苯醇抽提物含量和Klason木质素含量的近红外光谱分析模型。结果表明Klason木质素含量的预测效果优于苯醇抽提物含量的预测效果。然后在多模型方法的基础上,用预测误差较小的Klason木质素含量优化构建了苯醇抽提物含量的预测模型,使苯醇抽提物含量的预测效果得到改进。模型的拟合优度从0.792 8提升到0.827 1,预测值与实验值之间的相关系数从0.907 4提升到0.922 5。不同于已有的多模型方法,在优化建模时并不要求所使用的两种化学成分含量之间具有近似线性关系。还对优化构建的苯醇抽提物含量预测模型,通过减少每个子模型中待定常数的个数,增强了模型的稳定性,进一步改进了模型的预测效果。随着这方面研究工作的增多,未来该建模方法有希望应用于某些预测效果一般的化学成分含量,使这些化学成分含量的近红外光谱分析效果得到改进。

英文摘要：

Acacia is a kind of fast-growing wood pulp. The content of benzene alcohol extract has certain influence on pulp yield of wood. Near infrared spectra analysis can rapidly detect the contents of chemical components in wood with low cost. The multi-model method is a kind of near infrared spectra analysis modeling method, which has good prediction and is easy to master. The multi-model method has been applied to establish the near infrared spectral prediction model for detecting the contents of chemical components of acacia, populus tomentosa and populus euramericana, and better modeling effect is obtained. In the paper, the near infrared spectra analysis model for benzene alcohol extract of acacia contents and that of the Klason lignin of acacia were built by multi-model method at first. The results showed that the effect for predicting the contents of Klason lignin was better than that of the benzene alcohol extract. While, based on the multi-model method, the near infrared spectra analysis model for the benzene alcohol extract content was built optimally by the help of Klason lignin, whose prediction error was smaller than that of the content of benzene alcohol extract. It improved the prediction effect of the content of the benzene alcohol extract. The fit goodness of the model raised from 0.792 8 to 0.827 1. Different from the existing multi-model method, it was not required in this paper, when optimization modeling that the relationship between the contents of the 2 chemical components used to be approximately linear. For the prediction model of content of the benzene alcohol extract was built optimally, by reducing the number of undetermined constants in each sub model, the model stability was enhanced in the paper, and the prediction effect of this model was further improved. With the increase of research work in this area, this modeling method was hopeful for predicting the contents of chemical components, whose prediction effects were generic, and was also expected to improve the effects of near infrared s