光学学报, 2019, 39 (9): 0930004, 网络出版: 2019-09-09
山茶油中油酸和亚油酸近红外光谱分析模型 下载: 1169次
Analysis Model of Oleic and Linoleic Acids in Camellia Oil via Near-Infrared Spectroscopy
光谱学 近红外光谱 脂肪酸 变量筛选 蒙特卡罗无信息变量消除 变量组合集群分析 spectroscopy near-infrared spectroscopy fatty acid variable selection Monte Carlo uninformative variable elimination variable combination population analysis
摘要
将近红外光谱分析技术结合化学计量学方法用于山茶油混合油品中油酸和亚油酸含量的快速检测。配制了76种山茶油混合油样本用于近红外光谱的采集,将不同的光谱预处理方法用于光谱有效信息的提取;将蒙特卡罗无信息变量消除(MCUVE)和变量组合集群分析(VCPA)方法用于建模变量的选择;将偏最小二乘回归(PLSR)用于脂肪酸含量定量分析模型的构建。结果表明:经NWD1
st-MSC预处理后,两种脂肪酸的近红外光谱的较正均得到最好的结果;采用基于VCPA的变量优选方法极大地改善了模型精度,实现了建模变量数量的有效压缩。对于油酸模型,建模变量数量由1501减少为7,交叉验证均方根误差和校正相关系数分别为1.107和0.984,预测均方根误差和测试集的预测相关系数分别为1.178和0.981;对于亚油酸模型,建模变量数量由1501减少为8,交叉验证均方根误差和校正相关系数分别为0.089和0.987,预测均方根误差和测试集的预测相关系数分别为0.105和0.982。近红外光谱分析技术结合NWD1
st-MSC-VCPA-PLSR的方法为山茶油混合油品中脂肪酸含量的测定提供了一种快速简单的分析方法。
Abstract
Near-infrared spectroscopy (NIRS), combined with chemometrics methods, is applied to rapid quantitative determination of oleic acid and linolenic acid in camellia oil blends. 76 camellia oil samples are prepared and used for near-infrared spectral collection. Different spectral preprocessing methods are applied to effective information extraction. Two variable selection methods, Monte Carlo uninformative variable elimination (MCUVE) and variable combination population analysis (VCPA), are applied to select characteristic NIRS variables for the two fatty acids in camellia oil blends. Quantitative analysis models of the fatty acids are built using partial least-square regression. The results show that NWD1
st-MSC preprocessing can be used for optimization of near-infrared spectra of the two fatty acids in camellia oil blends. It is found that the VCPA method can greatly improve the precision of the model and compress the modeling variables. For the oleic acid model, the modeling variables decrease from 1501 to 7, the root-mean-square error of cross-validation and correlation coefficient of calibration are 1.107 and 0.984, respectively, and the root-mean-square error and correlation coefficient of prediction are 1.178 and 0.981, respectively. For the linoleic acid model, the modeling variables decrease from 1501 to 8, the root-mean-square error of cross-validation and correlation coefficient of calibration are 0.089 and 0.987, respectively, and the root-mean-square error and correlation coefficient of prediction are 0.105 and 0.982, respectively. NIRS combined with NWD1
st-MSC-VCPA-PLSR provides a quick and easy analysis method for measuring fatty acids in camellia oil blends.
郝勇, 吴文辉, 商庆园, 耿佩. 山茶油中油酸和亚油酸近红外光谱分析模型[J]. 光学学报, 2019, 39(9): 0930004. Yong Hao, Wenhui Wu, Qingyuan Shang, Pei Geng. Analysis Model of Oleic and Linoleic Acids in Camellia Oil via Near-Infrared Spectroscopy[J]. Acta Optica Sinica, 2019, 39(9): 0930004.