便携式近红外光谱仪与多种化学计量学方法快速鉴别西替利嗪片

Rapid qualitative analysis model for cetirizine hydrochloride Tablets by NIR using chemometric methods

采用近红外光谱法对不同厂家的盐酸西替利嗪片进行鉴别。用固体光纤在不同时间段采集了3个生产厂家的65批盐酸西替利嗪片的近红外漫反射光谱,对光谱数据进行预处理优化,采用基线校正,9点平滑,一阶导数和向量标准化的预处理方法,采用无监督学习算法即聚类分析法进行分类,并且比较了3种不同的聚类分析方法的分类结果;用有监督学习算法即人工神经网络法,运用改进的BP算法——Levenberg-Marquardt方法对46个样本建立校正模型,并且对其余的19个未知样本进行预测。聚类分析法和人工神经网络法都能得到满意的结果,其中经过主成分分析法提取特征变量后的聚类分析结果比直接进行聚类分析和经过核主成分分析法进行特征变量提取后的聚类分析的结果差。结果表明,用主成分分析法提取了前几个主成分不一定包含绝大部分聚类特征和结构,并且运用近红外光谱法与化学计量学结合可作为一种简单、快速、无损、可靠的方法用于鉴别不同厂家的西替利嗪片。

In this paper, the use of Near Infrared Spectroscopy （NIR） as a rapid and cost-effective classification techniques for cetirizine hydrochloride tablets have been preliminarily investigated. NIR spectra of 65 batches of cetirizine hydroehloride tablets that come from 3 manufacturers in China in the range of （12 000-3700） cm-l were recorded using optic fiber in different time. After pretreatment, unsupervised classification-clustering analysis and supervised classification-BP artificial neural network （ANN-BP） were adopted to classify, and unsupervised classification-clustering analysis included clustering analysis, principal component analysis （PCA）-clustering analysis and kernel principal component analysis （KPCA）-clustering analysis. This study compared the three clustering analysis methods, the correct recognition rates of two clustering analysis methods can reach 100% except principal component analysis （PCA）-elustering analysis. The result shows that the extracted principal components do not necessarily contain most of the clustering characteristic and structure using PCA. KPCA can effectively compute principal components in high dimensional feature spaces. It firstly maps the original inputs into high-dimensional feature space by non-linear mapping and then calculates principal components in the high-dimensional feature spaces, so KPCA overcomes the drawback of PCA which can only deal with linear problems; also, satisfactory results were obtained by supervised classifieation-ANN-BP. Overall, Near Infrared Spectroscopic method in the present paper played a good role in the discrimination, and offered a new approach to the rapid identification of eetirizine hydrochloride tablets.