基于光谱比值分析的无标样近红外模型传递方法
|
摘要
测量环境及光谱仪台间差异导致近红外光谱(NIRS)模型传递到从机后,常产生较大误差。该文使用标准正态变量变换(SNV)+微分处理光谱消除光谱散射和基线漂移的影响,提出通过仪器间光谱信号比值分析筛选波长的方法(Screening wavelengths based on spectrum ratio analysis,SWSRA),选出仪器间一致性较好且样本间差异大的光谱特征波长,采用筛选出的波长信号建立待测性质的偏最小二乘近红外光谱定标模型。以80个玉米样品中水分、油、蛋白质含量及72个黄芩样品中黄芩苷含量的NIRS预测对该方法进行了检验。结果表明,SWSRA主机模型预测从机样品的各成分含量的平均相对误差均小于4.3%,明显优于全波长模型直接传递的结果,且其预测均方根残差RMSEP与文献报道的其他模型传递方法的结果相当或更优。SWSRA方法具有模型参数少、稳健、简便易行等优点,可以在同类型近红外光谱仪器之间实现模型的无标样传递。
The changes of measurement environment and spectral signals of spectroscopy instruments could lead to big errors in the transferring process of the NIR(near infrared) spectra model to slave instruments. The influences of baseline drift and light scattering on NIR spectra were eliminated by standard normal variate transformation(SNV) plus derivative pretreatment in this paper. And then the wavelengths in which spectral signals are consistent and the spectral differences among samples are great were selected by screening wavelengths based on spectrum ratio analysis(SWSRA) among instruments. The partial least squares regression(PLS) models for predicting quality to be measured were built on the basis of the selected wavelength variables. Two data sets were utilized to validate the SWSRA method,in which one set was consisted of 72 Scutellaria baicalensis samples whose NIR spectra were measured on two types of NIR spectroscopy instruments,another one was composed of80 corn samples whose contents of moisture,protein and oils,and NIR spectra measured on three NIR instruments were available online. The results showed that the average errors of slave instruments for samples predicted by SWSRA models built on master instrument were all below 4. 3%. Root mean square error of prediction(RMSEP) of the method was much lower than that of the full-wavelength PLS model transferred to slaves,which was equivalent or superior to those of the other model transfer methods reported in literatures. With the advantages of few parameters,stability and simplicity,the SWSRA method could keep a good prediction performance during the transmission of spectral signals between the master and salve instruments. Similar near-infrared instruments could share the SWSRA models without standards.
The changes of measurement environment and spectral signals of spectroscopy instruments could lead to big errors in the transferring process of the NIR(near infrared) spectra model to slave instruments. The influences of baseline drift and light scattering on NIR spectra were eliminated by standard normal variate transformation(SNV) plus derivative pretreatment in this paper. And then the wavelengths in which spectral signals are consistent and the spectral differences among samples are great were selected by screening wavelengths based on spectrum ratio analysis(SWSRA) among instruments. The partial least squares regression(PLS) models for predicting quality to be measured were built on the basis of the selected wavelength variables. Two data sets were utilized to validate the SWSRA method,in which one set was consisted of 72 Scutellaria baicalensis samples whose NIR spectra were measured on two types of NIR spectroscopy instruments,another one was composed of80 corn samples whose contents of moisture,protein and oils,and NIR spectra measured on three NIR instruments were available online. The results showed that the average errors of slave instruments for samples predicted by SWSRA models built on master instrument were all below 4. 3%. Root mean square error of prediction(RMSEP) of the method was much lower than that of the full-wavelength PLS model transferred to slaves,which was equivalent or superior to those of the other model transfer methods reported in literatures. With the advantages of few parameters,stability and simplicity,the SWSRA method could keep a good prediction performance during the transmission of spectral signals between the master and salve instruments. Similar near-infrared instruments could share the SWSRA models without standards.
引文
[1]Chu X L.Molecular Spectroscopy Analytical Technology Combined with Chemometrics and Its Applications.Beijing:Chemical Industry Press(褚小立.化学计量学方法与分子光谱分析技术.化学工业出版社),2011:10-12.
[2]Ni L J,Luan S R,Zhang L G.Chin.J.Chin.Mater.Med.(倪力军,栾绍嵘,张立国.中国中药杂志),2016,41(19):3520-3527.
[3]Thygesen L G,Lundqvist S O.J.Near Infrared Spectrosc.,2000,8(3):183-189.
[4]Chu X L,Yuan H F,Wang Y B,Lu W Z.Spectrosc.Spectral Anal.(褚小立,袁洪福,王艳斌,陆婉珍.光谱学与光谱分析),2004,24(6):666-671.
[5]FerréJ,Feudale R N,Woody N A,Tan H,Myles A J,Brown S D.Chemom.Intell.Lab.Syst.,2002,64(2):181-193.
[6]Yu B,Ji H,Kang Y.Spectrosc.Lett.,2016,49(1):23-29.
[7]Zhang J,Cai W S,Shao X G.Prog.Chem.(张进,蔡文生,邵学广.化学进展),2017,29(8):902-910.
[8]Li X Y,Xu Z H,Cai W S,Shao X G.Anal.Chim.Acta,2015,880:26-31.
[9]Li X Y,Cai W S,Shao X G.J.Near Infrared Spectrosc.,2015,23(5):285-291.
[10]Zhang X Y,Li Q B,Zhang G J.Spectrosc.Spectral Anal.(张晓羽,李庆波,张广军.光谱学与光谱分析),2014,34(5):1429-1433.
[11]Wang J X,Li H,Xing Z N,Guo H G.J.Instrum.Anal.(王菊香,李华,邢志娜,郭恒光.分析测试学报),2011,30(1):43-47.
[12]Galv2o R K H,Araujo M C U,JoséG E,Pontes M J C,Silva E C,Saldanha T C B.Talanta,2005,67(4):736-740.
[13]Ni L J,Zhang L G.Basic Chemometrics and Its Applications.Shanghai:East China University of Science Press(倪力军,张立国.基础化学计量学及其应用.上海:华东理工大学出版社),2011:237.
[14]Luo Z C.Research on Autoencoder Modeling and Calibration Tranfer in Near-Infrared Spectroscopy.Guilin:Guilin University of Electronic Technology(雒志超.近红外光谱的自编码网络建模及模型传递方法研究.桂林:桂林电子科技大学),2015.
[15]Liu Y,Cai W S,Shao X G.Anal.Chim.Acta,2014,836:18-23.
[16]Boucher T,Dyar M D.J.Chemom.,2017,31(4):e2877.
[17]Chen W R,Bin J,Lu H M,Zhang Z M,Liang Y Z.Analyst,2016,141(6):1973-1980.
[18]Poerio D V,Brown S D.Appl.Spectrosc.,2018,72(3):378-391.
[2]Ni L J,Luan S R,Zhang L G.Chin.J.Chin.Mater.Med.(倪力军,栾绍嵘,张立国.中国中药杂志),2016,41(19):3520-3527.
[3]Thygesen L G,Lundqvist S O.J.Near Infrared Spectrosc.,2000,8(3):183-189.
[4]Chu X L,Yuan H F,Wang Y B,Lu W Z.Spectrosc.Spectral Anal.(褚小立,袁洪福,王艳斌,陆婉珍.光谱学与光谱分析),2004,24(6):666-671.
[5]FerréJ,Feudale R N,Woody N A,Tan H,Myles A J,Brown S D.Chemom.Intell.Lab.Syst.,2002,64(2):181-193.
[6]Yu B,Ji H,Kang Y.Spectrosc.Lett.,2016,49(1):23-29.
[7]Zhang J,Cai W S,Shao X G.Prog.Chem.(张进,蔡文生,邵学广.化学进展),2017,29(8):902-910.
[8]Li X Y,Xu Z H,Cai W S,Shao X G.Anal.Chim.Acta,2015,880:26-31.
[9]Li X Y,Cai W S,Shao X G.J.Near Infrared Spectrosc.,2015,23(5):285-291.
[10]Zhang X Y,Li Q B,Zhang G J.Spectrosc.Spectral Anal.(张晓羽,李庆波,张广军.光谱学与光谱分析),2014,34(5):1429-1433.
[11]Wang J X,Li H,Xing Z N,Guo H G.J.Instrum.Anal.(王菊香,李华,邢志娜,郭恒光.分析测试学报),2011,30(1):43-47.
[12]Galv2o R K H,Araujo M C U,JoséG E,Pontes M J C,Silva E C,Saldanha T C B.Talanta,2005,67(4):736-740.
[13]Ni L J,Zhang L G.Basic Chemometrics and Its Applications.Shanghai:East China University of Science Press(倪力军,张立国.基础化学计量学及其应用.上海:华东理工大学出版社),2011:237.
[14]Luo Z C.Research on Autoencoder Modeling and Calibration Tranfer in Near-Infrared Spectroscopy.Guilin:Guilin University of Electronic Technology(雒志超.近红外光谱的自编码网络建模及模型传递方法研究.桂林:桂林电子科技大学),2015.
[15]Liu Y,Cai W S,Shao X G.Anal.Chim.Acta,2014,836:18-23.
[16]Boucher T,Dyar M D.J.Chemom.,2017,31(4):e2877.
[17]Chen W R,Bin J,Lu H M,Zhang Z M,Liang Y Z.Analyst,2016,141(6):1973-1980.
[18]Poerio D V,Brown S D.Appl.Spectrosc.,2018,72(3):378-391.