摘要
应用便携式近红外光谱仪快速检测制浆材时,如果能实现同系列不同型号仪器之间分析模型共享,将极大降低仪器建模和维护成本。为实现混合木材木质素含量的近红外分析模型从1台主机向2台不同型号从机的模型传递,收集了5种常见制浆材的82个原木样品,经粉碎预处理后分别在3台便携式光谱仪上采集其近红外光谱信号,采用差谱、光谱的平均差异和光谱间的欧氏距离等方法,定量表征了仪器间的信号差异,分析并讨论了差异产生的原因。利用偏最小二乘回归建立了样品主机近红外光谱与木质素含量的关联模型,再分别采用斜率截距、直接校正和典型相关分析算法进行主机与两台从机间的模型传递,比较了模型传递前后预测精度。结果表明,便携式光谱仪间的差异多为非线性,且不同型号从机光谱仪间差异更为复杂。尽管主机向同型号的从机模型传递效果更优,但经直接校正算法和典型相关分析算法传递后两台不同型号从机预测相关系数均大于0.98、预测相对标准偏差均大于3、预测标准偏差均小于1.1%,可实现木材木质素含量的近红外光谱分析模型在3台便携式光谱仪间的传递。该研究结果对于不同型号便携式光谱仪分析模型共享具有重要意义。
Portable near infrared spectrometer is widely applied to determine the compositions of pulping materials.The cost of modeling and maintenance for the instrument will be greatly reduced if the analytical models can be shared among different types of instruments within the same series.The transfer of near-infrared calibration model for lignin determination from the main machine to other two different types of portable sub-unit spectrometers(slave)was investigated in this study.82 samples from 5 common wood pulps were collected and prepared by grinding pretreatment and their near-infrared spectrum signals were obtained via the three portable spectrometers.Three different methods,namely,difference spectra,average diversity of spectra and Euclidean distance between spectra were employed to quantitatively characterize the difference of spectrum signals between the three spectrometers.The reasons causing the differences between spectra from three portable spectrometers were also discussed.The near-infrared calibration model of lignin content in pulp samples was established via the master machine by partial least square regression.Then the established model was transferred from the master machine to the two slaves by algorithm of slope/bias(S/B),direct standardization(DS)and canonical correlation analysis(CCA).The accuracy of the model was evaluated by comparing the predicted value to that of the original model.The results indicated that the differences between portable spectrometers were mostly non-liner.And the difference between two types of slave spectrometers are more complicated.Although the transfer of calibration model within same type of instruments has the better performance,the prediction correlation coefficients of two different types of slaves by DS and CCA was up to 0.98,the ratio of performance to standard deviate was more than 3,and the root mean square error of prediction less than 1.1%.Therefore,the calibration model transfer between three portable near-infrared spectrometers could be achieved by both DS and CCA algorithm.This study could provide useful information for the near-infrared calibration model transfer among different types of portable spectrometers.
作者
刘耀瑶
熊智新
王勇
梁龙
房桂干
LIU Yaoyao;XIONG Zhixin;WANG Yong;LIANG Long;FANG Guigan(Jiangsu Provincial Key Lab of Pulp and Paper Science and Technology,Nanjing Forestry University,Nanjing 210037,China;Institution of Chemical Industry of Forestry Products,CAF,Nanjing 210042,China)
出处
《林业工程学报》
CSCD
北大核心
2019年第4期93-98,共6页
Journal of Forestry Engineering
基金
国家林业局948项目(2014-4-31)
关键词
近红外光谱分析
便携式光谱仪
模型传递
木质素
偏最小二乘回归
near-infrared spectroscopic analysis
portable spectrometer
model transfer
lignin
partial least square regression
