论文部分内容阅读
采用激光诱导击穿光谱(LIBS)技术对大豆油中的铬(Cr)含量进行检测研究。以一系列Cr含量不同的大豆油为样本,采用Ava Spec双通道高精度光谱仪在206.28~481.77 nm波段范围内采集LIBS光谱。根据样本的LIBS谱线图,确定Cr元素的主要特征谱线,并对Cr元素主要特征谱线应用线性回归或最小二乘支持向量机(LS-SVM)方法建立其单变量、二变量及多变量校正模型。利用建立的校正模型对样本Cr含量进行预测。研究结果表明,二变量及多变量校正模型的性能优于单变量校正模型,LS-SVM建立的多变量校正模型性能最优。对于单变量及二变量校正模型,预测样本的平均相对误差(RE)分别为14.16%和11.58%;而对于线性回归及LS-SVM建立的多变量校正模型,预测样本的平均RE分别为10.95%和4.97%。由此可见,LIBS技术检测大豆油中的重金属Cr含量具有一定的可行性,LS-SVM方法可以有效提高校正模型的预测精度。
Chromium (Cr) content in soybean oil was detected by laser induced breakdown spectroscopy (LIBS). Taking a series of soybean oil with different Cr content as sample, the LIBS spectra were collected by the Ava Spec dual-channel high-precision spectrometer in the range of 206.28 ~ 481.77 nm. According to the LIBS spectrum of the sample, the main characteristic lines of Cr element were determined, and the main characteristic lines of Cr element were established by LS-SVM or LS-SVM method to establish their univariate, Variable Correction Model. The calibration model was used to predict the Cr content of samples. The results show that the performance of the two-variable and multivariable calibration models is better than that of the univariate calibration model. The multivariable calibration model established by LS-SVM has the best performance. For the univariate and two-variable calibration models, the average relative errors (REs) of the predicted samples are 14.16% and 11.58%, respectively. For the multivariate calibration models established by linear regression and LS-SVM, the average REs of the predicted samples are 10.95% And 4.97% respectively. Thus, it is feasible to detect the content of heavy metal Cr in soybean oil by LIBS technique. The LS-SVM method can effectively improve the prediction accuracy of the calibration model.