本研究探讨模型传递方法在金银花醇沉过程中的适用性。在金银花醇沉过程中，以高效液相色谱法（HPLC）为对照分析方法，测定提取液中绿原酸的含量，分别运用iPLS算法、SPLS算法和模型更新（MU）建立绿原酸的NIR 定量校正模型，检测另外两个批次金银花醇沉过程中绿原酸的含量变化。为使选择的样本更具代表性，本研究在OSC的基础上提出了一种新的样本选择方法——正交空间样本选择（OS），并与常用的K-S法进行了对比。结果表明，采用模型更新可以实现定量校正模型在不同的批次之间稳健传递。且用OS法选择更新样本能够明显提高模型预测的准确性和稳健性。

In this article, orthogonal selection (OS) was proposed in the sample selection in order to update the raw calibration model. Models described the relationship between concentrations of chlorogenic acid. And the corresponding spectra were used to monitor the alcohol precipitation processes. A comparison between models built using model updating (MU) and interval partial least squares (iPLS). Stacked partial least squares (SPLS) was executed on three batches of Flos Lonicerae Japonicae extractions. It was found that MU can be used to successfully correct batch-wise variations. The performance of model updating with OS outperforms that of Kennard-Stone algorithm.

科学技术部国家“重大新药创制”科技重大专项（2010ZX09502-002）：清开灵注射液安全性关键技术研究，负责人：乔延江。