基于傅里叶变换衰减全反射红外光谱鉴别山茶油掺假

Identification of Camellia oil adulteration by attenuated total reflectance-Fourier transform infrared spectroscopy

中文关键词:  红外光谱  山茶油  掺假  判别分析

英文关键词:attenuated total reflectance-Fourier transform infrared spectroscopy  Camellia oil  adulteration  discriminant analysis

中文摘要:

      目的 基于傅里叶变换衰减全反射红外光谱(attenuated total reflectance-Fourier transform infrared spectroscopy, ATR-FTIR)结合聚类判别分析等化学计量学方法, 建立山茶油、大豆油、玉米油、葵花籽油和花生油5种植物油的快速鉴别模型, 及山茶油掺假模型。方法 采集山茶油、大豆油、玉米油、葵花籽油和花生油5种植物油共99份样品, 并按照不同质量百分比(掺伪5%~95%)将大豆油、葵花籽油、玉米油、1:1玉米大豆油、花生油、棕榈油掺入到山茶油中, 获得掺假山茶油样品196份, 采集600~4000 cm-1波段的红外光谱信息, 建立偏最小二乘判别分析(partial least squares-discriminant analysis, PLS-DA)、主成分分析-判别分析(principal component analysis-linear discriminant analysis, PCA-LDA)、K最近邻分类算法(K-nearest neighbor, KNN)以及数据驱动型簇类独立软模式分类(data driven soft independent modelling of class analogy, DD-SIMCA)模型, 并比较各方法建模效果, 确定最优识别模型。结果 各样品组红外吸收光谱非常类似, 具有相似的特征峰数、峰位置和峰形。DD-SIMCA建立的鉴别模型能将山茶油和其他类别植物油样本完全分开; PLS-DA、PCA-LDA和KNN模型判别经分析比较, 发现利用PLS-DA和PCA-LDA模型在5种植物油的分类中校正集和预测集中的各样本的预测值与实际值很接近, 除了花生油以外其余种类植物油的校正集和预测集样本的识别率和预测正确率均为100.0%; ATR-FTIR结合PLS的计量学方法能够准确进行山茶油掺假定量分析, 可用于掺杂大豆油、玉米油、葵花籽油等的定性定量分析, 结果可靠, 最低检出限可达5%。结论 ATR-FTIR结合聚类判别分析等化学计量学方法实现对山茶油掺假的高效识别。

英文摘要:

      Objective To establish a rapid identification model for 5 different types of vegetable oils (Camellia oil, soybean oil, corn oil, sunflower seed oil and peanut oil) and adulterated Camellia oil, using attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) and chemometrics methods such as cluster discriminant analysis. Methods The 99 samples of 5 different types of vegetable oils, including Camellia oil, soybean oil, corn oil, sunflower seed oil, and peanut oil were collected. According to the mass percentage of 5%–95%, soybean oil, sunflower seed oil, corn oil, 1:1 corn soybean oil, and palm oil was mixed into the Camellia oil, and 196 samples of the adulterated Camellia oil were obtained. Their infrared spectrum were collected in 600?4000 cm?1 region. The models for partial least squares discriminant analysis (PLS-DA), principal component analysis discriminant analysis (PCA-LDA), K-nearest neighbor (KNN), and data driven soft independent modeling of class analogy (DD-SIMCA) were established and compared to determine the optimal recognition model. Results The infrared spectra of each sample group had similar characteristic peaks, peak positions, and peak shapeswere with slight differences. The discriminant model established by DD-SIMCA could completely separate Camellia oil samples from those of other types of vegetable oil. By comparison of PLS-DA, PCA-LDA, and KNN models, it was found that the predicted values of each sample in the training and testing sets of the classification of 5 types of edible vegetable oils samples using PLS-DA and PCA-LDA models were accurate and reliable. Except for peanut oil, the recognition and prediction accuracy of the training and testing sets of other edible vegetable oils were both 100.0%. The quantitative analysis of Camellia oil adulteration using ATR-FTIR combined with PLS could be accurately carried out, which could be used for qualitative and quantitative analysis of adulterated soybean oil, corn oil, sunflower seed oil, etc. The results were reliable, and the lowest limit of detection could reach 5%. Conclusion Adulterated Camellia oil can be determined accurately and efficiently based on ATR-FTIR combined with chemometric methods.