文献类型：期刊文献

中文题名：SWCSS-DS算法在纸浆材综纤维素近红外分析模型传递中的应用

英文题名：Application of SWCSS-DS algorithm in transfer of near infrared analysis model for holocellulose in pulpwood

作者：王红鸿[1] 汪莹[1] 黄浩冉[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：2024

卷号：43

期号：6

起止页码：864-870

中文期刊名：分析试验室

外文期刊名：Chinese Journal of Analysis Laboratory

收录：CSTPCD;;北大核心:【北大核心2023】；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;DS;模型传递

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

分类号：O657.63

摘要：以纸浆材为研究对象,对其综纤维素含量模型进行传递分析,以解决不同近红外光谱仪间多元校正模型无法共享的问题。利用筛选出仪器间具有稳定一致光谱信号的波长(SWCSS)算法,选出稳定性较好的波长,并采用直接校正法(DS)对SWCSS方法校正后仍然存在的系统误差进一步校正。该联合算法能够提高主机模型的普适性,降低光谱矩阵维数,使模型转移更稳定和简单。将本文结果与单独的SWCSS, DS和分段直接标准化(PDS)算法校正后的传递结果进行比较。结果表明,与模型传递前分析能力相比,SWCSS-DS联用算法对2台从机样品的预测标准偏差(RMSEP)分别从2.48和2.31下降到了1.05和1.07,优于单独的SWCSS, DS和PDS算法结果。
In this paper,the holocellulose content model of pulpwood was transferred and analyzed to solve the problem that the multivariate calibration models between different near-infrared spectrometers could not be shared.The wavelength with stable and consistent spectral signals between the instruments(SWCSS)algorithm was used to select the wavelength with better stability,and the direct standardization(DS)method was used to further correct the systematic errors that still exist after being corrected by the SWCSS method.The combined algorithm can improve the universality of the master model,reduce the dimension of the spectral matrix,and make the model transfer more stable and simpler.The results of this paper were compared with the transfer results corrected by SWCSS,DS and piecewise direct standardization(PDS)algorithm.The results showed that the root mean squared error for prediction(RMSEP)of the SWCSS-DS combined algorithm for two slave samples decreased from 2.48 and 2.31 to 1.05 and 1.07,respectively,compared with the pre-transfer analysis ability of the model,which was better than the results of the single SWCSS,DS and PDS algorithms.