基于表面增强拉曼光谱(SERS)的青贮玉米原料农药残留快速检测方法研究

Rapid detection of pesticide residues in silage corn raw materials based on SERS

试验旨在建立1种基于表面增强拉曼光谱(SERS)技术检测青贮玉米原料中苯醚甲环唑、多菌灵、啶酰菌胺农药残留的方法。采用快速萃取净化的前处理方法实现了青贮玉米原料中3种农药的提取,以10 mg/kg氯化钠溶液作为SERS增强剂结合化学物质浓缩法获取了青贮玉米原料中不同农药残留的SERS信号,结合Savitzky-Golay卷积平滑法、二阶导数法、自适应迭代重加权惩罚最小二乘法(airPLS)以及SG-airPLS进行光谱数据的预处理,融合了线性判别式分析(LDA)-支持向量机(SVM)、卷积神经网络(CNN)两种算法建立了针对苯醚甲环唑、多菌灵、啶酰菌胺的识别模型,并对模型进行对比分析。结果表明,拉曼峰1 127、1 229、1 237 cm-1分别作为识别苯醚甲环唑、多菌灵和啶酰菌胺的拉曼特征峰,当青贮玉米原料中的3种农药残留含量分别为0.1 mg/kg时,仍清晰可见各农药相应的特征峰。采用SG-airPLS方法预处理后的LDA-SVM分类模型可以识别所有样本,准确度为86.67%。研究表明,SERS和SG-airPLS-LDA-SVM方法的融合可以准确、快速识别青贮玉米原料中的农药成分。

The study was to establish a qualitative method for determination of difenoconazole,carbendazim and acetanilide pesticide residues in silage corn based on SERS.Pretreatment method the rapid extraction and purification was used to extract three pesticides from silage corn raw material.The SERS signals of different pesticide residues in silage corn raw material were obtained by using 10 mg/kg sodium chloride solution as SERS enhancer combined with chemical substance concentration method.The spectral data were preprocessed by combining Savitzky-Golay convolution smoothing method,second derivative method,airPLS and SG-airPLS.The LDA-SVM and CNN algorithms were integrated to establish a recognition model for phenoxymethyclozole,carbendazim and dinimide.The results showed that the Raman peaks 1127,1229 and 1237 cm^(-1)were used as Raman fingerprint peaks to identify difenoconazole,carbendazim and acetanilide,respectively.When the residues of three pesticides in silage corn raw materials were 0.1 mg/kg,the corresponding characteristic peaks of each pesticide were still clearly visible.The LDA-SVM classification model pretreated by SG air PLS could recognize all samples with an accuracy of 86.67%.The study indicates that the fusion of SERS and SG air PLS-LDA-SVM methods can accurately and quickly identify the pesticide components in silage corn raw materials.