文献类型：期刊文献

中文题名：基于稳定一致波长优选的综纤维素近红外分析模型传递

英文题名：Model transfer for near infrared analysis of holocellulose based on stable and consistent wavelength optimization

作者：王红鸿[1] 熊智新[1] 胡云超[1] 刘智健[1] 梁龙[2]

第一作者：王红鸿

机构：[1]南京林业大学轻工与食品学院,南京210037;[2]中国林业科学研究院林产化学工业研究所,南京210042

年份：2023

卷号：42

期号：11

起止页码：1468-1475

中文期刊名：分析试验室

外文期刊名：Chinese Journal of Analysis Laboratory

收录：CSTPCD;;EI(收录号：20234915171644);北大核心:【北大核心2020】；CSCD:【CSCD_E2023_2024】；

基金：中国林科院林业新技术所基本科研业务费专项(CAFYBB2019SY039)资助。

语种：中文

中文关键词：综纤维素含量;近红外光谱;稳定一致波长;波长优选;模型传递

外文关键词：holocellulose content;near infrared spectroscopy;stable consistent wavelength;wavelength optimization;model transfer

分类号：O657.3

摘要：以实现纸浆材综纤维素含量的近红外分析模型在3台不同型号光谱仪上共享为目标,提出SWCSS-UVE及SWCSS-CARS联用算法。即分别利用竞争性自适应重加权采样算法(CARS)和无信息变量剔除(UVE)算法,减少SWCSS方法中入选的无信息或信息少波长的不利影响,以提高模型转移精度,并与单独的SWCSS和分段直接标准化算法(PDS)以及斜率截距(S/B)算法校正后的传递结果进行比较。结果表明,通过SWCSS-UVE方法最终可从稳定一致光谱信号中进一步优选出91个波长建立模型,该模型能同时应用于2台从机所测量光谱的分析,预测标准偏差(RMSEP)分别从模型转移前的2.0114和9.4518下降到了1.5919与1.6818,优于SWCSS,SWCSS-CARS和PDS以及S/B算法的结果。这表明SWCSS-UVE算法可以有效剔除SWCSS方法中包含的无效波长,简化模型传递过程,提高模型传递效率和稳健性。
The combined algorithm of SWCSS-UVE and SWCSS-CARS was proposed to realize the aim of achieving the sharing of the near infrared analysis model of the holocellulose content of pulp material on three different spectral instruments.Competitive adaptive reweighted sampling(CARS)and uninformative variables elimination(UVE)algorithm were used to reduce the adverse effects of wavelengths with no information or less information selected in the SWCSS method to improve model transfer accuracy.The results were compared with the transfer results corrected by the separate SWCSS,the piecewise direct standardization algorithm(PDS)and the slope intercept(S/B)algorithm.And the results showed that the SWCSS-UVE method can finally establish a model by selecting 91 wavelengths from the stable and consistent spectral signal.The method can be applied in the analysis of the spectra measured by the two instruments at the same time.The root mean square error of prediction(RMSEP)dropped from 2.0114 and 9.4518 to 1.5919 and 1.6818 respectively compared to before model transfer,which was superior to the results of SWCSS,SWCSS-CARS,PDS and S/B algorithm.The results showed that the SWCSS-UVE algorithm can effectively eliminate the invalid wavelength contained in the SWCSS method,greatly simplify the model transfer process and improve the efficiency and robustness of the model transfer.