摘要
为了提高木质素近红外分析模型转移的质量,提出了将直接标准化(DS)与分段直接标准化(PDS)联用的DS-PDS方法,即利用PDS算法校正使用DS方法全局校正后仪器之间仍然存在的局部差异,并以3台光谱仪(棱光为主机、 Si-ware和Insion为从机)测得的81个样品木质素光谱数据为研究对象,验证该方法的可行性。结果表明,该联用算法能充分发挥DS算法与PDS算法各自的优点,与单独使用DS的最优结果相比,标样集并未增加,PDS算法的窗口宽度、偏最小二乘回归(PLSR)模型的主因子数均减小,对从机样品的预测精度进一步提高。对Si-Ware而言,相对标准偏差(RPD)值由DS方法的3.3853提高到4.0335, Insion的RPD值则从2.8600提高到3.4462。DS-PDS联用算法除了使用更少的参数,还能减弱模型转移中光谱不连续现象,弥补了单独使用DS和PDS算法的缺点,为近红外光谱的模型转移提供了一种新的解决方案。
In order to improve the transfer performance of the lignin near-infrared analysis model,a method combining direct standardization(DS)and piecewise direct standardization(PDS),named DS-PDS,was proposed.The PDS algorithm was used to correct the local differences between instruments still existed after the global correction using the DS method.The lignin spectral data of 81 samples measured by three spectrometers(Lengguang as master instrument,and Si-Ware and Insion as target instruments,respectively)were used to verify the feasibility of the new method.The experimental results showed that the combined algorithm can exploit the advantages of DS algorithm and PDS algorithm respectively.Compared with the optimal results of using DS algorithm alone,the standard sample set did not increase.For PDS algorithm,the window width,and the number of principal factors of the partial least square regression(PLSR)model were reduced,while the prediction accuracy of the target instruments was further improved.The ratio of prediction to deviation(RPD)value for Si-Ware increased from 3.3853 to 4.0335 comparing to the previous DS method,and the RPD value for Insion increased from 2.8600 to 3.4462.In addition to using fewer parameters,the combined DS-PDS algorithm can also reduce the phenomenon of spectral discontinuity in model transfer,which makes up for the shortcomings of using DS and PDS algorithms alone,and provides a new solution to the model transfer strategy of near-infrared spectroscopy.
作者
刘智健
熊智新
胡云超
汪莹
黄浩冉
王红鸿
梁龙
LIU Zhijian;XIONG Zhixin;HU Yunchao;WANG Ying;HUANG Haoran;WANG Honghong;Liang long(College of Light Industry and Food,Nanjing Forestry University,Nanjing 210037,China;Institute of Chemical Industry of Forest Products,Chinese Academy of Forestry,Nanjing 210042,China)
出处
《分析试验室》
EI
CAS
CSCD
北大核心
2024年第4期502-510,共9页
Chinese Journal of Analysis Laboratory
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。
