波长筛选结合直接校正法用于近红外光谱模型传递研究  

Application of Wavelength Selection-Direct Standardization Algorithm to Calibration Transfer of Near Infrared Analysis

作　　者：王菊香[1] 李华[2] 邢志娜[1] 郭恒光[2] 

机构地区：[1]海军航空工程学院飞行器工程系,山东烟台264001 [2]海军航空工程学院研究生管理大队,山东烟台264001

出　　处：《分析测试学报》2011年第1期43-47,共5页Journal of Instrumental Analysis

基　　金：国防预研基金资助项目(9140A19050106JB1409)

摘　　要：该文提出了一种新的传递方法模型———波长筛选结合直接校正法(WSDS)。首先利用样品的性质信息筛选出最有代表性的波长点信息,然后用直接校正法消除这些信息中包含的仪器间差异,以预测标准偏差(SEP)考察模型传递的效果。利用此算法对航空煤油的近红外光谱分析模型在不同仪器之间进行传递研究。经WSDS校正后,对航空煤油密度预测的SEP由0.004 6降至0.001 2,对航空煤油20℃时粘度预测的SEP由0.058 6降至0.020 9。结果表明,该算法能消除仪器间的大部分差异,较好地实现模型传递效果。与直接校正法(DS)、分段直接校正法(PDS)的比较表明,WSDS不仅在精度上能取得比DS、PDS更好的效果,而且能有效地对光谱数据进行压缩,是一种集压缩、光谱校正于一体的模型传递新方法。In this paper, a wavelength selection -direct standardization （WSDS） algorithm was proposed for calibration transfer of NIR model. The WSDS is one of the modifications of direct standardization（DS）. In WSDS, only the wavelengths which were highly correlated with the sample property were selected for correcting the variations of the instrumental responses of two different NIR spectrometers. The corrected spectroscopic data were then used to build the model for analysis. The application of WSDS algorithm to calibration transfer of NIR analysis was evaluated by analyzing the jet fuel with different NIR spectrometers. The results showed that after the WSDS correction, the squared error of prediction（SEP） of the density of jet fuel decreased from 0. 004 6 to 0. 001 2 and that of the viscosity（20 ℃ ） of jet fuel decreased from 0. 058 6 to 0. 020 9. The results showed that this method can well correct the difference between instruments and improve the analytical accuracy. The corrected results, obtained by WSDS, direct standardization （DS） and piecewise direct standardization （PDS） , respectively, for jet fuel with different NIR spectrometers indicated that the former gave the best analytical accuracy.

关 键 词：近红外 模型传递 波长筛选 直接校正法 航空煤油 

分 类 号：O657.33[理学—分析化学]