基于傅里叶变换红外光谱的大豆原油掺伪定量分析模型研究

摘要

Objective To establish a rapid method for discrimination of adulterated oils from crude soybean oils by Fourier transform infrared spectroscopy (FTIR) combined with partial least squares method (PLS). Methods The spectrograms of 42 crude soybean oils, 21 refined edible oils and 88 adulterated oils were collected and analyzed for the establishment of adulteration model. After preprocessing these spectra data with standard normal variate (SNV), principal component analysis (PCA) was used to extract the characteristic variables in pattern recognition. And PLS discrimination model was established by randomly picking 60 adulterated oils as calibration set, and 28 adulterated oils as validation set. Results The PCA results showed that the crude soybean oils and refined oils can be divide into separate categories and the classification limit of palm oil mixed with crude soybean oil was 5%. The calibration model of PLS was well suited for quantitative analysis of palm oil with the coefficient of determination (R2) of 0.9926 and the root mean standard error of calibration (RMSEC) of 1.8121. Prediction models with R2 and the root mean standard error of cross validation (RMSECV) were 0.9823 and 2.8189 respectively. While the deviation of predicted results obtained were between 1.3909%~3.1019%, the difference was not significant, and the results illustrated calibration model was feasible. Conclusion FTIR-PLS model can discriminate crude soybean oil adulteration with palm oil, and FTIR is rapid and suitable for field testing, it will provide a rapid and convenient method for crude soybean oil adulteration analysis.%目的：应用傅里叶变换红外光谱(FTIR)结合最小偏二乘法(PLS)建立大豆原油-棕榈油二元掺伪体系的定量分析模型。方法以42个大豆原油、21个精炼油、88个掺伪油的FIIR谱图为模型样本,预处理方法选用标准正态变量(SNV),在此基础上应用主成分分析(PCA)提取特征变量,随机选取60个掺伪油样组成校正集,28个掺伪油样组成验证集,以PLS方法建立大豆原油的掺伪定量模型。结果 PCA可将大豆原油及精炼油分成独立的2类。经PCA分析,大豆原油中掺入棕榈油的掺伪检测限为5%。PLS校正模型的判定系数R2为0.9926,校正误差均方根RMSEC为1.8121。预测模型的R2为0.9823,交叉验证误差均方根RMSECV为2.8189。同时得到的预测结果的偏差在1.3909%～3.1019%之间,差异不显著,说明此模型可行。结论 FTIR-PLS模型能够实现大豆原油的掺伪定量分析,分析速度快,能够满足大豆原油入库要求,是一种可行的大豆原油掺伪分析方法。