Research On Quantitative Analysis Method Of Cr Elements In Potato Based On LIBS

Through research,it has been shown that the main heavy metal pollutants in potatoes are lead(Pb),cadmium(Cd),and chromium(Cr).Of these elements,chromium is particularly unique as it is not only an essential trace element for the growth and development of potatoes,but also one of the main elements responsible for causing heavy metal pollution.Therefore,In this paper,potatoes are used as the object of research to detect the content of the heavy metal element Cr.Laser-induced breakdown spectroscopy(LIBS)is a rapid method for the detection of elements in substances.Quantitative and qualitative analysis are two important areas of application for LIBS technology.This article mainly studies the application of LIBS technology in quantitative analysis.Among them,the focus of quantitative analysis includes two aspects:one is the contribution rate of important element characteristic spectral lines to quantitative analysis results,and the other is the accuracy of quantitative analysis results.Therefore,this article starts from two key research aspects of quantitative analysis and studies the analytical methods used to calculate the Cr element content in potatoes.The main research content is as follows:(1)The potato spectral data were collected using the LIBS detection platform and subjected to pre-processing operations such as noise reduction and baseline correction.The quantitative analysis of Cr element in potatoes is carried out using the calibration curve method,and the calibration curve is drawn.The predictive performance is evaluated using model evaluation indicators R~2,RMSEC,RMSEP and ARE.The results show that the R~2value of the calibration curve model reaches 0.9109,while RMSEC,RMSEP,and ARE are0.00399wt.%,0.00381wt.%,and 18.85%,respectively.Overall,the calibration curve method produces results with higher error and lower accuracy.(2)Due to the influence of matrix elements such as Ca and Na on the characteristic spectral lines of Cr elements in potato,the accuracy of single line calibration results is not high.Therefore,in order to further improve the accuracy of quantification,the BP neural network and Bagging ensemble learning multispectral analysis method are introduced.Compared with the model evaluation index results of the absolute strength method,the R~2value of the BP neural network model is obtained,which is 0.986.RMSEC,RMSEP,and ARE decreased to 0.00159wt.%,0.00253wt.%,and 11.98%,respectively.The results show that the quantitative accuracy of Cr element in the BP neural network model is significantly improved.However,it can be found through the evaluation index results of the model that the average relative error of the BP neural network model is relatively large,reaching over10%.This may be due to issues such as local extremum and poor stability in BP neural networks.Therefore,in response to the problems encountered in training potato data using the BP neural network,the Bagging ensemble learning algorithm is used to further optimize the BP neural network.After optimization,the values of the evaluation indicators RMSEC,RMSEP,and ARE decrease to 0.00093wt.%,0.00115wt.%,and 5.39%,respectively.This indicates that Bagging ensemble learning effectively optimizes the BP neural network,further improving quantitative accuracy.(3)Due to the fact that the BP neural network is trained with multiple feature spectral lines as input variables,these feature spectral lines are not only limited to multiple Cr element spectral lines,but also matrix element feature spectral lines such as Ca and Na,and some other feature spectral line data with less obvious peaks.However,the Bagging integrated BP neural network cannot see the contribution rate of each feature spectral line to the high-precision quantitative analysis results,making it difficult to determine whether the high-precision quantitative results are determined by the Cr element feature spectral line.This leads to unclear sources of high-precision quantitative analysis results,making it difficult to explain.Therefore,two important representative models of Boosting integrated decision trees,GBDT and XGBoost,are introduced to analyze the contribution rate of the characteristic lines of important elements in potato to the quantitative analysis results.The results show that XGBoost has higher quantitative accuracy compared to other quantitative analysis methods,with R~2,RMSEC,RMSEP,and ARE of 0.9999,0.00030wt.%,0.00113wt.%,and 4.26%,respectively.By analyzing the contribution rate of characteristic spectral lines,it is found that the total contribution of Cr element characteristic spectral lines in the XGBoost model results is as high as 0.58898,an increase of 0.06805 compared to the GBDT model.In addition,the contribution rate of other matrix element characteristic spectral lines is relatively low,which further verifies that the XGBoost model can effectively reduce the impact of inter element interference and make the source of high-precision quantitative results clearer and clearer.