利用激光诱导击穿光谱(LIBS)技术对大豆油中的重金属Cr进行检测研究。 以松木木片对重金属Cr进行富集, 采用AvaSpec双通道高精度光谱仪在206.28~481.77 nm波段范围内采集松木木片样本的LIBS光谱, 利用无信息变量消除(UVE)方法筛选与重金属Cr相关的波长变量, 应用偏最小二乘(PLS)回归建立大豆油中重金属Cr的定标模型, 并与单变量及全波段PLS定标模型进行比较。 结果表明, 相比单变量及全波段PLS定标模型, UVE-PLS定标模型的性能更优, 其相关系数、 校正均方根误差、 交互验证均方根误差及预测均方根误差分别为0.990, 0.045, 0.050及0.054 mg·g-1。 经UVE变量筛选后, UVE-PLS定标模型所用的波长变量数仅为全波段PLS的2%。 由此可见, UVE是一种有效的波长变量筛选方法, 能有效筛选出与重金属Cr相关的波长变量。

In order to monitor chromium (Cr) content in soybean oil, laser induced breakdown spectroscopy (LIBS) was used to detect Cr content in this research. Pine wood chips was used to enrich heavy metal of Cr, and the spectra of pine wood chips were acquired in the wavelength range of 206.28~481.77 nm by a two-channel high-precision spectrometer. Then, uninformative variable elimination (UVE) method was used to select sensitive wavelength variables for heavy metal of Cr, and calibration model of Cr in soybean oil was developed with partial least squares (PLS) regression, the performance of the calibration model was compared to univariate and full PLS calibration models. The results indicate that the performance of UVE-PLS calibration model is better than that of univariate and full PLS calibration models, the correlation coefficient, root mean square error of calibration (RMSEC), root mean square error of cross validation (RMSECV), root mean square error of prediction (RMSEP) are 0.990, 0.045 mg·g-1, 0.050 mg·g-1 and 0.054 mg·g-1, respectively. After UVE variable selection, the number of wavelength variables in UVE-PLS calibration model is about 2% of wavelength variables in full PLS calibration model. This means UVE is an effective variable selection method which can select correlative variables for heavy metal of Cr.