DS算法在近红外光谱多元校正模型传递中的应用

Application of DS Algorithm to the Calibration Transfer in Near-Infrared Spectroscopy

模型传递问题是近红外光谱分析技术中一个关键的共性基础技术问题。模型传递通过在同一型号的两台仪器之间寻求一种变换关系,使得在一台仪器上建立起的模型能够应用于另外一台仪器样品光谱响应的预测。文章将直接校正(direct standardization,DS)算法应用于化学计量学多元校正中的模型传递,并研究了模型转换集样品的选择方法。在两台AOTF近红外光谱仪上进行了模型传递实验,首先采用Kennard/Stone算法选择转换集样品,然后采用DS方法进行模型传递。实验结果表明采用DS算法取得了较好的模型传递效果。DS算法不仅可以应用于不同仪器之间的光谱分析模型传递,而且该方法对于同一仪器的长时间漂移或者由于部件的更换、测量环境的改变等引入的光谱差异校正也是适用的。

Calibration transfer in multivariate calibration is one of the most important and key issues in near-infrared spectral analysis technology. The model was transferred by means of finding the transformation relation between two instruments of the same type, so that the model established on one instrument could be used on the other to predict the spectral response. In this paper, direct standardization （DS） algorithm was investigated for resolving calibration transfer in multivariate calibration of chemometrics. In addition, the selection method for the transfer set samples was studied in this paper. The experiment of model transfer was carried out between two AOTF near-infrared spectrometers. Firstly, the Kennard/Stone algorithm was adopted to select the transfer set samples, and then DS algorithm was applied to the calibration transfer. The experimental results showed that DS algorithm was effective for model transfer. DS algorithm is applicable to not only instrument standardization, but also the correction of the spectral differences induced by baseline drift for a long time or the replacement of measurement modules for the same instrument.