基于激光诱导击穿光谱技术的烟草重金属检测研究

发布时间：2018-05-27 13:35

本文选题：激光诱导击穿光谱技术 + 烟草　；参考：《浙江大学》2017年硕士论文

【摘要】：近年来,由人为活动造成的重金属污染使植物体内积累过量的重金属元素,重金属超标的水果、蔬菜、粮食等农产品会对人体健康造成危害。烟草是我国重要经济作物之一,重金属污染不仅影响香烟品质,而且对吸烟者的健康产生严重的危害。因此,烟草中重金属元素检测对香烟品质管理及吸烟者身体健康有着重要意义。本论文以重金属(铜Cu和镉Cd)胁迫下的烟草为研究对象,研究烟草重金属污染程度判别及重金属元素定量检测的方法和模型,具体研究内容和结果如下:(1)研究了烟草鲜叶Cu污染程度LIBS判别方法。在土培条件下,采摘不同程度Cu污染下的烟草叶片并收集其LIBS光谱信号。对全谱LIBS数据进行主成分分析,基于PC1、PC2、PC3的载荷值选取对应于C、Fe、Mg、Ca、Fe、S、N、O等元素的18条谱线作为特征变量。利用所选特征变量建立线性判别分析(LDA)和支持向量机(SVM)分类模型。在LDA模型中,建模集与预测集的识别正确率分别为85%和72%;在SVM模型中,建模集与预测集的分类正确率分别为97%和82%。结果表明,基于LIBS技术结合化学计量学方法判别新鲜烟草叶片Cu污染程度是可行的,能获得满意的判断结果。(2)研究基于响应面法的烟草重金属定量分析LIBS仪器参数优化方法。利用BBD响应面法针对LIBS检测烟草重金属含量的三个仪器因素:①激光脉冲能量,②探测器相对于激光脉冲的延时时间,③探测器门宽(积分时间),设计了实验方案,确定了最佳的三因素组合。a.在烟草叶样本Cu含量检测的参数优化实验中,以Cu I:324.75 nm谱线的信背比(Cu I:324.75 nm谱线强度与325.67 nm～326.73 mn范围光谱平均值之比)作为目标函数,以最大化目标函数为目标分析实验结果,最终确定基于LIBS技术的烟草Cu定量分析的参数优化组合为:激光能量为109 mJ,延时时间为5.19μs,积分时间为14.66μs。在此条件下得到的最大信背比为20.8857。b.在烟草根样本Cd定量分析的参数优化实验中,以Cd I:361.05 nm谱线的信背比(Cd I:361.05 nm谱线强度与359.15 nm～360.45 nm范围光谱平均值之比)作为目标函数,以使目标函数最大化为目的分析实验结果,确定基于LIBS技术的烟草Cd定量分析的参数优化组合为:激光能量为115.34 mJ,延时时间为4.41μs,积分时间为6.48μs。此参数组合下最大信背比为11.7614。(3)研究建立了烟草重金属元素LIBS快速定量检测方法和模型。①在基于LIBS烟草叶样本Cu含量分析中,选择对应Cu I的324.75 nm和327.4 nm谱线作为关键变量。利用上述关键变量分别建立基于单变量的一元线性回归模型和基于双变量的MLR模型。此外,利用320 nm～330 nm范围LIBS光谱数据建立PLS模型。基于Cu I 324.75 nm谱线的一元线性回归模型中,建模集决定系数R_c~2为0.96,预测集决定系数R_p~2为0.94;基于Cu I 327.4 nm谱线的一元线性回归模型中,R_c~2为0.96,R_p~2为0.91;基于Cu I的324.75 nm和327.4 nm谱线的MLR模型中,R_c~2为0.96,Rp/为0.92;基于320 nm～330 nm范围光谱的PLS模型中,R_c~2为0.95,Rp/为0.90。②在基于LIBS烟草根样本Cd含量分析中,选择对应CdI的346.62 nm和361.05 nm谱线作为关键变量。利用上述关键变量分别建立基于单变量的一元线性回归模型和基于双变量的MLR模型。此外,利用340 nm～365 nm范围LIBBS光谱数据建立PLS模型。基于Cd I 346.62 nm谱线的一元线性回归模型中,R_c~2为0.96,R_p~2为0.90;基于CdI 361.05 nm谱线的一元线性回归模型中,R_c~2为0.96,R_p~2为0.92;基于CdI 346.62 nm和361.05 nm谱线的MLR模型中,R_c~2为0.97,R_p~2为0.92;基于340 nm～365 nm范围光谱的PLS模型中,R_c~2为0.98,R_p~2为0.94。结果表明,基于LIBS技术的烟草重金属定量分析可行,基于单变量的建模结果与基于多变量的建模结果相近,为后续便携式LIBS植物重金属检测仪器的开发提供了理论条件。
[Abstract]:In recent years, heavy metal pollution caused by human activities has accumulated excessive heavy metal elements in plants, and agricultural products, such as fruits, vegetables and grain, will be harmful to human health. Tobacco is one of the most important economic crops in China. Heavy metal pollution not only affects the quality of cigarettes, but also produces serious health for smokers. Therefore, the detection of heavy metal elements in tobacco is of great significance to the quality management of cigarettes and the health of smokers. In this paper, the study of tobacco under the stress of heavy metals (copper Cu and CD Cd) is the research object, and the methods and models of heavy metal pollution degree discrimination and quantitative determination of heavy metals are studied. The specific contents and results are as follows (1) the LIBS discrimination method of Cu pollution degree of fresh leaves of tobacco was studied. Under the condition of soil culture, tobacco leaves under different degree of Cu pollution were picked and their LIBS spectral signals were collected. The total spectral LIBS data were analyzed by principal component analysis. The load values of PC1, PC2, PC3 were selected as the characteristic variables of C, Fe, Mg, Ca, Ca, PC3 and other elements. The classification model of linear discriminant analysis (LDA) and support vector machine (SVM) is established by the selected feature variables. In the LDA model, the recognition accuracy of the modeling set and the prediction set is 85% and 72% respectively. In the SVM model, the classification accuracy of the modeling set and the prediction set is 97% and 82%. respectively, and the new method is based on the combination of LIBS and chemometrics methods. The degree of Cu pollution in fresh tobacco leaves is feasible and can obtain satisfactory results. (2) study on the parameter optimization method of quantitative analysis of heavy metals in tobacco based on response surface method and three instrumental factors for detecting heavy metal content of tobacco using BBD response surface method: (1) excitation pulse energy and second detector relative to laser pulse. The time delay, the width of the detector gate (integral time), the experimental scheme is designed, and the best three factor combination.A. is determined in the parameter optimization experiment of Cu content detection in tobacco leaf sample. The target function is the ratio of the signal back ratio of the Cu I:324.75 nm line (the ratio of the spectral line strength of Cu I:324.75 nm and the spectral average of the 325.67 nm to 326.73 Mn range) as the target function. The maximum target function is the result of the target analysis. Finally, the parameters optimization combination of the quantitative analysis of tobacco Cu based on LIBS technology is as follows: the laser energy is 109 mJ, the time delay time is 5.19 s, the integral time is 14.66 U S., the maximum signal back ratio is the parameter optimization experiment of Cd quantitative analysis of 20.8857.b. in the tobacco root sample. In the case of Cd I:361.05 nm spectrum line signal back ratio (the ratio of Cd I:361.05 nm spectral line strength and 359.15 nm to 360.45 nm range spectral mean) as the target function, the objective function maximization is analyzed, and the optimization combination of quantitative analysis of tobacco Cd based on LIBS technology is determined as: laser energy is 115.34 mJ, time delay time The method and model for rapid quantitative detection of heavy metal elements LIBS in tobacco were established for 4.41 Mu s and 6.48 S. with the maximum signal back ratio of 11.7614. (3). (1) 324.75 nm and 327.4 nm spectrum lines corresponding to Cu I were selected as the key variables in the Cu content analysis of LIBS tobacco leaves. A univariate linear regression model based on a single variable and a bivariate MLR model are established. In addition, the PLS model is established using 320 nm ~ 330 nm range LIBS spectral data. In the linear regression model based on the Cu I 324.75 nm line, the modeling set determination coefficient R_c~2 is 0.96, the prediction set coefficient R_p~2 is 0.94, and the Cu I 327.4 spectral line is based. In a linear regression model, R_c~2 is 0.96 and R_p~2 is 0.91; R_c~2 is 0.96 and Rp/ is 0.92 in the MLR model of 324.75 nm and 327.4 nm lines based on Cu I, and 0.95 in PLS model based on 320 nm ~ 330 nm spectrum. Line is a key variable. Using the above key variables, one variable linear regression model and MLR model based on bivariate are established respectively. In addition, PLS model is established by using 340 nm ~ 365 nm range LIBBS spectral data. In the linear regression model based on Cd I 346.62 nm line, R_c~2 is 0.96, R_p~2 is 0.90; CdI 361.05 n is based. In the linear regression model of M line, R_c~2 is 0.96 and R_p~2 is 0.92; R_c~2 is 0.97, R_p~2 is 0.92 in MLR model based on CdI 346.62 nm and 361.05 nm spectrum line, and 0.98 in PLS model based on 340 nm ~ 365 nm range spectrum. The modeling results are similar to those based on multivariate modeling, which provides a theoretical basis for the subsequent development of portable LIBS plant heavy metal detection instruments.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS411

【相似文献】