共轴双脉冲激光诱导击穿光谱和最小二乘支持向量机法定量分析植物油中铬

利用共轴双脉冲激光诱导击穿光谱( DP-LIBS)技术对植物油(大豆油、花生油和玉米油)中的重金属铬( Cr)含量进行定量分析。采用Ava-Spec双通道高精度光谱仪采集样品的LIBS光谱,然后通过其LIBS谱线图确定了CN分子谱线(421.49 nm)、Ca原子谱线(422.64 nm)及Cr的3条原子谱线(425.39、427.43和428.87 nm),根据上述谱线建立了Cr元素的单变量定标模型和最小二乘支持向量机(LS-SVM)校正模型,并用验证样品对它们进行检验。研究结果表明,对于单变量定标法,大豆油、花生油及玉米油验证样品的平均预测相对误差(PRE)分别为12.57%,12.11%和13.72%；对于三变量LS-SVM法,其定标样品真实值与预测值之间的拟合度 R2分别为0.9785,0.9792和0.9654,验证样品的平均 PRE 分别为8.92%,8.33%和10.98%；对于五变量LS-SVM法(增加两基体元素谱线变量),其定标样品真实值与预测值之间的拟合度R2分别为0.9895,0.9901和0.9855,验证样品的平均PRE分别为7.46%,8.96%和8.95%。由此可知,LS-SVM校正模型性能优于单变量定标法,且五变量LS-SVM校正模型性能优于三变量LS-SVM校正模型；采用LS-SVM法及引入合适的基体元素谱线( CN、Ca)能有效减小定量分析误差,提高LIBS技术对植物油中Cr含量预测的精度。

Quantitative Analysis of Chromium in Vegetable Oil Based on Double Pulse-Laser-induced Breakdown Spectroscopy and Least Squares Support Vector Machine

The content of heavy metal chromium ( Cr) in vegetable oil ( soybean oil, peanut oil and corn oil) was quantitatively analyzed by collinear double pulse laser-induced breakdown spectroscopy ( DP-LIBS ) . An Ava-Spec two-channel spectrometer was used to acquire LIBS spectra of samples, and then the CN molecular spectral line (421. 49 nm), Ca atomic spectral line (422. 64 nm) and three atomic lines of Cr (425. 39 nm, 427. 43nm and 428. 87 nm) were determined by LIBS spectrum of samples. According to the above spectral lines, single variable calibration model and least squares support vector machine ( LS-SVM) calibration model of Cr element were established, and they were verified by the validation samples. The results showed that the average prediction relative errors ( PRE) of the validation samples by using single variable calibration method were 12. 57%, 12. 11% and 13. 72%, respectively. When using LS-SVM method of three variables, the fitting degree (R2) between the true value and the predictive value of the calibration samples was 0. 9785, 0 . 9792 and 0 . 9654 , and the average PRE of the validation samples was 8 . 92%, 8 . 33% and 10 . 98%, respectively. While using LS-SVM method of five variables, the fitting degree ( R2 ) between the true value and the predictive value of the calibration samples was 0 . 9895 , 0 . 9901 and 0 . 9855 , and the average PRE of the validation samples was 7. 46%, 8. 96% and 8. 95%, respectively. Therefore, the LS-SVM calibration model has better performance than single variable calibration method. And the LS-SVM calibration model established with five variables has better performance than the one with three variables. The LS-SVM method and the introduction of suitable matrix elements ( CN, Ca ) can reduce the error of quantitative analysis effectively, and improve the accuracy of prediction for Cr content in vegetable oil by LIBS technique.