基于偏最小二乘法的磷脂红外快速检测

Rapid Test of Phospholipids by Infrared Spectroscopy Combined with Partial Least-Square Regression

本试验介绍一种采用傅里叶变换红外光谱仪(FTIR)结合偏最小二乘法快速检测植物油中磷脂含量的方法。采集了400～4 000 cm-1波数范围内的磷脂红外光谱,并采用偏最小二乘法(PLS)对磷脂含量在200～20 000 mg/kg的植物油样品建立定量分析模型。结果表明,采用SNV的预处理方法,在970～1 320cm-1波数范围内建立的定量分析模型,其相关系数(R2)达到0.997 9,校正均方差(RMSEC)和预测均方差(RMSEP)分别为316,386。同传统的钼蓝比色法相比,FTIR/PLS法测得的磷脂含量的相对标准偏差为1.42%～3.52%,平均相对偏差为2.36%,均优于钼蓝比色法的结果,两者平均相对误差为0.75%,说明FTIR/PLS法具有良好的精密度和准确度,能够实现植物油中磷脂的快速检测,并有简单、高效的优点。

A rapid reliable method was developed to determine the total phospholipids content in vegetable oils by Fourier transform - mid infrared spectroscopy （ FT - MIR） and partial least - square regression （PLS）. Infrared spectroscopy in wave range of 400 - 4 000 cm - 1 was collected, and quantitative analysis model of phospholipids con- tent in 200 - 20 000 mg/kg vegetable oil samples was established by using the PLS regression. The results showed that, in the quantitative analysis model established in wave range of 970- 1 320 cm- 1 by using standard normalization of vector （SNV） preprocessing method, the correlation coefficient reached 0. 997 9, and the RMSEC （316） and RMSEP （386） values were also determined. Compare with traditional Molybdenum blue method, the relative stand- ard deviation of determined phospholipids content with FTIR/PLS method is 1.42% - 3.52%, the relative mean de- viation is 2.36%, both higher than the results of Molybdenum blue method; the mean relative error is 0.75%, which showed good precision and accuracy of models with FTIR/PLS method, realizing the rapid test of phospholipids in vegetable oil with simple and efficient characteristics.