红外光谱快速测定油茶籽油脂肪酸组成的模型建立Establishment of rapid detection model of fatty acid composition of oil-tea camellia seed oil by infrared spectroscopy

Abstract

通过气相色谱和傅里叶红外光谱仪测定86个油茶籽油样本的脂肪酸组成和红外光谱图，采用支持向量机（SVM）和BP人工神经网络（ANN）的非线性建模方法，构建油茶籽油中主要脂肪酸的定量回归模型。结果表明：ANN建立的油酸和棕榈酸定量回归模型精确度比SVM高，校正集的相关系数（R）分别为0.998 7和0.945 1，预测集的相关系数分别为0.955 7和0.926 2，相对标准偏差分别小于1%和5%；SVM和ANN建立的亚油酸定量分析模型精确度都非常高，相对标准偏差均小于1%。说明红外光谱用于油茶籽油中主要脂肪酸的快速检测是完全可行的。 The fatty acid composition and infrared spectra of 86 kinds of oil-tea camellia seed oil samples were determined by gas chromatography and Fourier transform infrared spectrometer, and the nonlinear modeling methods of support vector machine (SVM) and BP artificial neural network (ANN) were used to construct the quantitative regression model of main fatty acids in oil-tea camellia seed oil. The results showed that the quantitative regression models of oleic acid and palmitic acid established by ANN were more accurate than those by SVM, the correlation coefficients(R) of the correction set and the prediction set were 0.998 7, 0.945 1 and 0.955 7, 0.926 2, respectively, and the relative standard deviations were less than 1% and 5%, respectively. The accuracies of linoleic acid quantitative analysis models established by SVM and ANN were both very high, and the relative standard deviation was both less than 1%. It showed that infrared spectroscopy was feasible for the rapid detection of main fatty acids in oil-tea camellia seed oil.

Keywords

• 油茶籽油；脂肪酸组成；红外光谱；支持向量机；人工神经网络；模型
• oil-tea camellia seed oil; fatty acid composition; infrared spectroscopy; support vector machine; artificial neural network; model