近红外光谱定量校正模型适用性研究

Study of Quantitative Calibration Model Suitability in Near infrared Spectroscopy Analysis

近红外光谱作为一种依靠模型进行分析的技术 ,对测定的样品进行模型适用性判断是得到可靠分析结果的前提。对于通过校正集样本近红外光谱测量和标准方法测定的基础数据依靠因子分析技术建立的多元校正模型 ,提出将因子分析与Mahalanobis距离相结合判断定量模型适用性的方法 ,以近红外光谱测定柴油十六烷值为例 ,对影响模型适用性判断的一些因素进行了讨论。

Near-infrared spectroscopy is a fast, high efficiency and low cost analytical technique, that depended on analytical models which are built through near-infrared spectra, the primary method results of calibration set and chemometrics methods. The reliability of analytical results mostly depends on the suitability of analytical model to unknown samples. To judge the suitability of analytical model, a new method is presented that combine principle factor analysis (PCA) and Mahalanobis distance. It is different with traditional method, that the Mahalanobis distance is calculated through PCA factor scores. It can avoid the wavelengths selection problem in traditional method of spectral Mahalanobis distance. Through the determination of cetane number of diesel fuel, some factors that affect model suitability judgement are discussed.