Study Of Rapid Detection Method And Detection System Based On Electrochemical Technology For Pesticide Residues In Agricultural Products

Irrational use and abuse of pesticides in the planting process of agricultural products lead to serious pesticide residues in agricultural products.The classical chemical detection method has high sensitivity,but the pre-treatment is complex and costly,requiring professional operation.Electrochemical detection technology has the advantages of fast detection speed,high sensitivity,easy integration and miniaturization,widely used for rapid detection of quality and safety.In this paper,the electrochemical detection methods for pesticide residues in agricultural products were investigated,combined with cyclic voltammetry（CV）and differential pulse voltammetry（DPV）,and a set of portable electrochemical detection system based on smart phones was developed.The main research contents are as follows:（1）A laser-induced porous graphene electrochemical sensor was construct and machine learning methods were investigated for rapidly analyzing maleyl hydrazine residues in potatoes and peanuts.A computer-controlled micromachining system controlled by the computer was used to directly work on polyimide films,to construct a flexible electrode known as laser-induced porous graphene（LIPG）.And based on the electrochemical sensing platform,maleyl hydrazine（MH）residue in potatoes and peanuts was analyzed using machine learning（ML）method.The chemical properties of LIPG electrode showed that the electrode had high conductivity,large effective surface area and good stability.Maleic hydrazide concentration range was divided into six different range of concentration data,and artificial neural network（BP-ANN）,random forests（RF）and least squares support vector machines（LS-SVM）method were used to develop the prediction models for analying the maleic hydrazide pesticide residues in potatoes and peanuts and compare with one another,to explore the optimum pesticide concentration range predicted by the developed models.The results showed that the prediction performance of three models was the best in the concentration range of 0.9～101.9μM,and the LS-SVM model was the best among the three models with the minimum prediction error.And in the training set,R² of the model was 0.9670,and the error RMSE and MAE were 6.4095 and 4.6415,respectively,and in the prediction set,R² of the modelwas 0.9804,the error RMSE and MAE were 4.6187 and 3.6581,and the RPD was 6.7721.The average recovery rate of MH in potatoes and peanuts was 97.96～106.44%,and RSD was 1.12～1.56%,indicating that the LS-SVM model was feasible for MH residue.（2）MWCNTs/GO/AgNPs nanoenzyme sensor was constructed and machine learning method was used to rapidly analyze benendazim residues in tea and cucumbers.Dendritic AgNPs were electrodeposited on the surface of MWCNTs/GO by electrochemical deposition method to prepare nanocomposites and an electrochemical sensor based on MWCNTs/GO/AgNPs was constructed.And machine learning（ML）method was used to investigate benendazim（BN）residues in tea leaves and cucumbers.Orthogonal experiment method combined with neural network（BP-ANN）and genetic algorithm（GA）was used to optimize four experiment parameters:the volume ratio of MWCNTs-GO,concentration of AgNO₃,deposition cycle number and p H of buffer solution.And the optimal experimental conditions were as follows:the volume ratio of GO:MWCNTs was 1,the concentration of AgNO3 was 19,the number of CV deposition cycles was 17,and the p H value of PBS solution was 5.8.The chemical properties of the prepared MWCNTs/GO/AgNPs nano-enzyme sensor showed that the electrode had large effective surface area,high electrocatalytic capacity,good stability and anti-interference.Support vector machine（SVM）and least square support vector machine（LS-SVM）were used to construct prediction models for analyzing benendazim residue,and compared with traditional linear regression.The results showed that LS-SVM model had the best performance in predicting benendazim residue,and the prediction error was the least.The R²,RMSE,MAE of the training set samples were 0.9997,0.1306 and 0.0894,respectively.The R²,RMSE,MAE and RPD of the prediction set samples were 0.9963,0.6023,0.3272 and 10.2,respectively.The average recovery of BN in tea and cucumber was 97.05%～101.28%,and RSD was 0.1%～1.75%,indicating that MWCNTs/GO/AgNPs nanocomposites could be used for the determination of BN pesticide residues in agricultural products.（3）Portable electrochemical detection system combined with screen printing electrode was used for rapid analysis of carbendazim residues in tea and rice by.ZSHPC modified layered porous carbon（ZSHPC）was mixed with MWCNTs to modify screen-printed electrode,and a portable electrochemical detection system and machine learning method were used to investigate carbendazim（CBZ）residues in rice and tea.The chemical properties of the prepared electrode showed that the electrode had good electrocatalytic ability,large effective surface area,strong stability and anti-interference ability.Back Propagaton Neural Network（BP-ANN）、Support Vector Machine（SVM）and Least square support vector machine（LS-SVM）model was used to predict carbendazim residues.Compared with the traditional linear regression,the LS-SVM model had the best performance and the lowest prediction error.The R²,RMSE,MAE of the training set samples were 0.9969,0.3605 and0.2968,respectively.The R²,RMSE,MAE and RPD of the prediction set samples were0.9924,0.6190,0.5360 and 10.3097,respectively.The average recovery of CBZ in tea and rice was 98.77%～109.32%.RSD was 0.47%～2.58%,indicating that the rapid analysis of carbendazim pesticide residues in agricultural products based on portable detection system combined with machine learning was feasible.（4）Development of portable electrochemical detection system based on smart phones.Based on the electrochemical detection method ideas established above,a set of portable electrochemical detection system based on smart phones was developed using Java language on the Linux platform.The functions of cyclic voltammetry（CV）and differential pulse voltammetry（DPV）method selection,parameters design,Bluetooth function connection,data preprocessing,peak current peak search and real-time display of voltammetry curve were realized.The detection system could fit the curve of the data by the least square method,and correct the specific points.The curve was displayed in the user interface and stored in real time.The detection system could run on a mobile browser and was connected to a portable electrochemical workstation via Bluetooth.The developed system had strong compatibility and could reduce memory footprint.