应用透反射技术在1 100~2 500 nm波谱段采集大豆油近红外光谱, 采用改进的偏最小二乘法算法, 建立了近红外光谱快速测定大豆油中五种主要脂肪酸含量的方法。 以气相色谱法测定的158个大豆油样品中棕榈酸(C16∶0)、 硬脂酸(C18∶0)、 油酸(C18∶1)、 亚油酸(C18∶2)和亚麻酸(C18∶3)含量作为其化学值, 建模集样品数为138, 检验集样品数为10, 盲样验证集样品数为10; 通过对定标模型的优化, 五种脂肪酸的交互定标决定系数(1-VR)分别为0.883 9, 0.583 0, 0.900 1, 0.977 6, 0.959 6, 交互定标标准误差(SECV)分别为0.42, 0.29, 0.83, 0.46, 0.21; 盲样验证集样品五种脂肪酸的近红外预测值与化学值的相对标准误差均小于5.50%。 结果表明, 近红外预测值与化学值之间存在较好的线性关系, 所建立的方法快速、 方便、 可靠, 可用于大豆油的掺伪鉴别。

In the present research, a novel method was established for determination of five fatty acids in soybean oil by transmission reflection-near infrared spectroscopy. The optimum conditions of mathematics model of five components(C16∶0, C18∶0 , C18∶1, C18∶2 and C18∶3) were studied, including the sample set selection, chemical value analysis, the detection methods and condition. Chemical value was analyzed by gas chromatography. One hundred fifty eight samples were selected, 138 for modeling set, 10 for testing set and 10 for unknown sample set. All samples were placed in sample pools and scanned by transmission reflection-near infrared spectrum after sonicleaning for 10 minute. The 1 100~2 500 nm spectral region was analyzed. The acquisition interval was 2 nm. Modified partial least square method was chosen for calibration mode creating. Result demonstrated that the 1-VR of five fatty acids between the reference value of the modeling sample set and the near infrared spectrum predictive value were 0.883 9, 0.583 0, 0.900 1, 0.977 6 and 0.959 6, respectively. And the SECV of five fatty acids between the reference value of the modeling sample set and the near infrared spectrum predictive value were 0.42, 0.29, 0.83, 0.46 and 0.21, respectively. The standard error of the calibration (SECV) of five fatty acids between the reference value of testing sample set and the near infrared spectrum predictive value were 0.891, 0.790, 0.900, 0.976 and 0.942, respectively. It was proved that the near infrared spectrum predictive value was linear with chemical value and the mathematical model established for fatty acids of soybean oil was feasible. For validation, 10 unknown samples were selected for analysis by near infrared spectrum. The result demonstrated that the relative standard deviation between predict value and chemical value was less than 5.50%. That was to say that transmission reflection-near infrared spectroscopy had a good veracity in analysis of fatty acids of soybean oil.