Detection Of Diquat And Procymidone In Solution Based On Double Pulse Laser Induced Breakdown Spectroscopy And Variable Screening Algorithm

In recent years,the safety of pesticide residues in agricultural products has attracted more and more attention.Pesticide can not only control harmful organisms but also be one of the important factors to maintain the sustainable development of agriculture.It is an important means to ensure the stability,high yield and quality of agriculture.However,while pesticides bring high products to humans,it also bring some certain side effects to human health and even to the ecological environment.In the process of using pesticides,they are inevitably flew into the water,more seriously,pesticides would tend to leave in the environment during the spraying.So that the pesticide may accumulate on plants and animals,and then do harm to human’s health through foods.The worst result is to threaten human life.This requires researchers to test the environmental behavior of pesticides.In this paper,eutectic double pulse LIBS technology was used to study the content of diquat and procymidone in xylene and edible oil,and quantitative analysis was carried out by combining different variable screening methods.The main research contents and results are as follows:Laser-induced breakdown spectroscopy(LIBS)technology has the advantages of fast and reality simultaneous detection of multiple elements,so applying double-pulse LIBS technology to detect the content of diquat and procymidone.In order to configure different concentrations of diquat and procymidone samples,mix diquat and procymidonein a certain concentration with xylene or a mixture of xylene and oil dissolved completely in different proportions.Since liquid samples are easy to spill liquid during laser striking,has a certain of risk.Therefore,in this experiment,the liquid sample was converted into a solid sample,the graphite was used to adsorb the pesticide solution,and then the LIBS spectrum of the sample was collected by an eight-channel high-precision spectrometer,and the spectral data was pretreated by different pretreatment methods.In order to improve the detection accuracy of procymidone,the two-channel spectral data with the strongest signal of bromine and chlorine are selected,and using normalization,baseline correction,and standard normal variable are used respectively multiplicative scatter correction(MSC)method for spectral preprocessing,and applying PLS method for modeling.After comparing the data of each preprocessing method,select the optimal preprocessing method.Based on the optimal preprocessing method,uninformed variable elimination(UVE),competitive adaptive reweighted sampling(CARS)algorithm,genetic algorithm(GA)and the least squares support vector machine(LSSVM)eliminates the wavelength variable,screens important wavelength variables related to diquat and procymidone,and finally uses partial least squares regression to establish diquat and procymidone in solution.In the quantitative analysis of the rapid content of diquat in xylene,the GA-CARS method is preferred and the LSSVM method is used to model the calibration set and prediction set determination coefficients of 0.94 and 0.97,respectively,and the RMSEC and RMSEP are respectively1.6g/kg,and 1.1g/kg;in the quantitative analysis of the content of diquat in edible oil,after the GA-CARS method is preferred,the correction set and prediction set determination coefficients of the model are 0.93 and 0.94,respectively,and the RMSEC and RMSEP are respectively 1.6g/kg and 1.8g/kg;for the quantitative analysis of procymidone content in xylene,after the UVE-CARS method is preferred,the calibration set and prediction set determination coefficients of the model are 0.9970 and0.9920,respectively,and the RMSEC and RMSEP are respectively 0.35g/kg and0.60g/kg;for the quantitative analysis of the rapid content of diquat in edible oil,after the UVE-CARS method is preferred,the correction set and prediction set determination coefficients of the PLS model are 0.9909 and 0.9747,respectively,and the RMSEC and RMSEP arerespectively 0.62 g/kg and 1.13 g/kg.The modeling results show that the model performance is better than the PLS model established by the original spectrum after the spectral preprocessing and variable screening algorithms are optimized.The results show that it is feasible to quantitatively detect the content of diquat and procymidone in solution by using the coaxial double pulse LIBS technique.After screening by the variable screening algorithm,the characteristic variables and related influence variables can be effectively extracted,and the redundancy and noise are eliminated influenced variables,simplifying the quantitative analysis model and improving the stability of the quantitative analysis model.