基于近红外光谱的煎炸油极性组分定量分析模型构建

为了建立煎炸油中极性组分的快速检测方法，通过热处理和模拟煎炸方式，采集氧化程度不同的油样并用 国标法分析样品极性组分，采集样品近红外透射光谱，经光谱预处理，利用偏最小二乘法建立煎炸油极性组分定量 分析模型并对模型进行验证。结果表明：在波长范围为4 963～4 616、5 2 22～5 037 cm－1和5 688～5 499 cm－1，采 用一阶求导和Savitzky-Golay（7,5）平滑光谱处理，校正集相关系数为0.996 5，校正均方根差为1.84%，验证集R为 0.993 6，验证均方根差为1.92%，模型预测效果良好，利用近红外透射光谱测定煎炸油极性组分可行。

Quantitative Analysis Calibration of Polar Components of Frying Oils Based on Near Infrared Spectroscopy

This study established a quick method for determining the polar components of frying oil. The oil samples with
different degrees of oxidation through heating and simulated frying were collected and their polar components were analyzed
according to the Chinese national standard method. Near infrared (NIR) transmission spectra of these samples were recorded
and preprocessed, and a quantitative analysis model was developed using partial least square regression (PLSR) method and
validated. In the selected spectra ranges of 4 963–4 616, 5 222–5 037 and 5 688–5 499 cm-1, after spectral preprocessing
by first derivative and Savitzky-Golay (7,5) smoothing, the correlation coefficient (R) for the calibration set and the root
mean square error of calibration (RMSEC) were determined to be 0.996 5 and 1.84%, respectively, and the R value for the
validation set and the root mean square error of validation (RMSEV) were 0.993 6 and 1.92%, respectively, suggesting
the accuracy and reliability of the predictive model. Therefore, it is feasible to apply NIR to quantitatively analyze polar
components of frying oil.