正交信号校正用于傅里叶变换红外光谱的模型传递

Orthogonal Signal Correction Used for Calibration Transfer of Fourier Transform Infrared Spectra

利用正交信号校正(OSC)实现了4组分气体混合物的PLS模型,在两台傅里叶变换红外光谱(FT-IR)仪上的传递,并与直接标准化(DS)、分段直接标准化(PDS)、多元分散校正(MSC)和有限脉冲响应(FIR)的传递效果进行了比较,确立了以源机的校正模型直接对目标机数据进行预测的传递方式。经过OSC校正,预测均方根误差(RMSEP)为10-3左右。OSC可以有效地减小测量仪器间的差异,同时使PLS模型的潜变量个数降为4,使模型简单化。与DS和PDS相比,OSC不需要同一样品在两台仪器上测量,在支集大小为3时,得到一致的预测准确度,表现出稳健性。MSC和FIR处理的RMSEP为10-1左右,效果远差于OSC。

Orthogonal signal correction （OSC） was used for the calibration transfer of partial least squares （PLS） modal in which four-component gas mixtures were analyzed simultaneously between two fourier trasform infrared （FTIR） instruments. Suitable transfer process of OSC, calibration model built with data of one instrument was used directly for the prediction of the data from the second instrument, was determined to obtain good prediction accuracy with the root mean squared error of prediction（RMSEP） of about 10 ^-3. The transfer results with OSC were compared with those gotten by DS（ direct standardization）, PDS（ piecewise direct standardization）, MSC （multiplicative scatter correction） and FIR（ finite impulse response）. OSC was characterized of both model simplicity with 4 latent variables of PLS model and robustness with the subset size of 3. RMSEP of DS and PDS Were about 10^-3 which was consistent with the one of OSC but it＇s unnecessary to measure the same samples on different instruments for OSC. The RMSEP of MSC and FIR were about 10^-1, inferior to OSC greatly.