应用共轴双脉冲激光诱导击穿光谱(DP-LIBS)技术对食用植物油中重金属Cr含量进行快速定量检测。采用二通道高精度光谱仪采集样品的激光诱导击穿光谱(LIBS),根据LIBS在420~430 nm波段范围确定Cr元素的三条原子谱线 (Cr I 425.39 nm、Cr I 427.43 nm、Cr I 428.87 nm)、CN分子谱线 (CN 421.49 nm)及Ca原子谱线 (Ca II 422.64 nm);然后利用竞争性自适应重加权采样(CARS)方法筛选Cr元素的特征变量及相关影响变量,并应用最小二乘支持向量机(LSSVM) 建立Cr含量的定标模型。结果表明:经CARS方法优选后,波长变量个数由 132个减少为10个,变量压缩率为92.42%;CARS-LSSVM定标模型的相关系数、校正均方根误差及预测均方根误差分别为0.9926、5.287×10 -6和5.860×10 -6,预测集样品的平均相对误差为8.55%,优于单变量及五变量LSSVM定标模型。DP-LIBS技术定量检测食用植物油中的Cr含量具有一定的可行性,CARS方法可以有效筛选Cr元素的特征变量及相关影响变量,剔除冗余及噪声变量,从而有效降低了基体效应对分析元素的影响,提高了LIBS分析的预测精度。

Collinear double pulse laser induced breakdown spectroscopy (DP-LIBS) is applied to quickly and quantificationally detect the content of heavy metal chromium (Cr) in edible vegetable oil. LIBS spectra of samples are collected by a two-channel high precision spectrometer, and then several spectral lines such as three atomic lines of Cr (Cr I 425.39 nm, Cr I 427.43 nm, Cr I 428.87 nm), CN molecular line (CN 421.49 nm) and Ca atomic line (Ca II 422.64 nm) are determined at wavelength range of 420-430 nm. Then, competitive adaptive reweighted sampling (CARS) method is used to select characteristic and related variables of Cr, and least squares support vector machine (LSSVM) method is used to establish calibration model using selected variables. The results show that the number of variables reduces from 132 to 10 after CARS variable selection, and the variable compression rate is 92.42%. The correlation coefficient, root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) in CARS-LSSVM calibration model are 0.9926, 5.287×10 -6 and 5.860×10 -6, respectively, and the relative error of prediction samples is 8.55%. The performance of CARS-LSSVM calibration model is superior to that of univariate calibration model and LSSVM calibration model with five variables. So it can be concluded that DP-LIBS technique is feasible to detect the content of Cr in edible vegetable oil, and CARS method can select characteristic and related variables of Cr effectively, eliminate redundant and noise variables, thus reduce the influence of matrix effect on analytical element and improve prediction accuracy of LIBS analysis.