论文部分内容阅读
为了探索利用激光诱导击穿光谱(LIBS)对水田污染区稻壳中铬(Cr)元素含量进行绿色、快速检测的可行性,采用LIBS结合联合区间偏最小二乘法(SiPLS),对产自江西省某湖周边24个水田污染区稻壳样品中的Cr元素进行了定量分析。利用原子吸收光谱法(AAS)测得样品中Cr元素的真实浓度为32.51~510.33μg/g,利用LIBS光谱获得的Cr元素三个特征谱线Cr I 425.43nm、Cr I 427.48nm和Cr I 428.97nm清晰明显。对稻壳样品在422~446nm波段的LIBS光谱数据进行九点平滑处理后,在采用SiPLS获得的最佳模型基础上,得出模型交叉验证均方根误差与预测均方根误差分别为26.1μg/g和22.6μg/g,训练集相关系数与预测集相关系数分别为0.9714和0.9840。对预测集样品进行相对误差及T检验分析,结果显示稻壳中Cr元素浓度的预测值与AAS法测量的真实值之间的平均相对误差为6.17%,且无显著性差异,表明模型具有较好的预测精度,可为自然条件下生长的农产品重金属安全绿色分析提供参考依据。
In order to explore the feasibility of green and rapid detection of chromium (Cr) in rice husk by laser-induced breakdown spectroscopy (LIBS), LIBS and combined interval partial least squares (SiPLS) Cr elements in the rice husk samples from 24 paddy-contaminated areas around a lake in Hunan province were quantitatively analyzed. The true concentration of Cr in the sample measured by atomic absorption spectrometry (AAS) was 32.51 ~ 510.33μg / g. The three characteristic lines Cr I 425.43nm, Cr I 427.48nm and Cr I 428.97 nm clear and clear. Based on the best model obtained from SiPLS after smoothing the LIBS spectral data of rice husk samples in the band of 422 ~ 446 nm by nine points, the root mean square error of the cross validation and the prediction root mean square error of the model were 26.1 μg / g and 22.6μg / g respectively. The correlation coefficients between training set and prediction set were 0.9714 and 0.9840, respectively. The relative error and the T-test analysis of the predictive set samples showed that the average relative error between the predicted value of Cr elemental concentration in rice hulls and the true value measured by AAS method was 6.17%, and there was no significant difference, indicating that the model has more Good prediction accuracy can provide a reference for the green analysis of heavy metal safety of agricultural products grown under natural conditions.