为实现混合体系中多组分有机磷农药的定性识别和定量检测, 采用紫外-可见吸收光谱法结合平行因子分析法(PARAFAC), 对水体中多组分有机磷农药混合溶液进行快速分析测定。 在纯净水中配置毒死蜱、 甲基对硫磷、 丙溴磷的单组分、 2组分和3组分农药溶液为实验样本, 采用紫外-可见光谱仪获取各组样本的吸收光谱。 将测得的纯净水-有机磷农药吸收光谱数据构建为不同的三维数据矩阵, 经核一致诊断法确定因子数后采用PARAFAC算法对三维数据进行分解, 结果发现2组分和3组分混合农药经分解后得到的光谱图与实际单组分光谱图相似度很高, 表明算法可以实现水体中多组分有机磷农药的定性分析。 进一步地, 利用算法分解得到的得分矩阵与各组分的真实浓度构建线性回归模型, 再对不同的数据集(包括农田水为稀释背景的光谱数据集)进行预测。 模型的预测结果表明, PARAFAC算法具有显著的二阶校正优势, 即使是光谱重叠严重、 预测集中存在校正集中不存在的干扰信息, 算法依然可以有效地从混合体系中检测出目标物农药。 模型对2组分混合溶液均实现了定性分析与定量检测, 预测集决定系数R2都大于0.9, 预测残差RPD也都大于3; 对3组分混合溶液的毒死蜱、 甲基对硫磷、 丙溴磷实现了定性分析, 其中毒死蜱和甲基对硫磷均达到了定量检测要求, 只有丙溴磷的定量检测结果不理想, 可能是由于丙溴磷溶液的整体光谱强度水平显著低于相同浓度的毒死蜱和甲基对硫磷溶液, 其光谱贡献最小, 以致算法对其混合体系中丙溴磷的分辨效果较差。 PARAFAC算法实现了“数学分离”代替“化学分离”的效果, 不需要复杂的预处理即可对光谱重叠严重的多组分有机磷农药混合物进行定性识别和定量检测, 这一方法为水体中有机磷农药残留的快速检测分析提供了理论依据。

In order to realize the qualitative identification and quantitative detection of multi-component organophosphorus pesticides in mixed systems, this paper combines ultraviolet-visible absorption spectra with Parallel Factor Analysis (PARAFAC) to analyze the mixed solution of multi-component organophosphorus pesticides in water rapidly. The absorption spectra of experimental samples of single-component, two-component and three-component pesticide solutions composed of chlorpyrifos, methyl-parathion and profenofos in pure water were obtained by UV-Vis spectrometer. These pure water-organophosphorus pesticide absorption spectrum data were constructed into different three-dimensional data matrices. Then the PARAFAC algorithm was used to decompose the three-dimensional data after the factor number was determined by the nuclear consensus diagnosis method. It was found that the spectrum obtained by the decomposition of two-component and three-component pesticides was very similar to the actual single-component spectrum, which shows the algorithm can realize the qualitative analysis of multi-component organophosphorus pesticides in water. A linear regression model was constructed using the score matrix obtained by the algorithm decomposition and the true concentration of each component to predict different data sets (including a spectral data set with farmland water as dilution background). The prediction results of the model show that the PARAFAC algorithm has a significant second-order advantage. Even when the spectral overlap is serious, and there is interference information in the prediction set that does not exist in the calibration set, the algorithm can still effectively detect the mixed system. Qualitative analysis and quantitative detection were achieved for all the two-component mixed solutions, with the model evaluation coefficient of R2 greater than 0.9 and the RPD greater than 3. The qualitative analysis was achieved for chlorpyrifos, methyl-parathion, and propamocarb in the three-component mixed solutions, in which chlorpyrifos and methyl-parathion met the quantitative detection requirements, and only profenofos showed unsatisfactory quantitative detection results. It may be that the overall spectral intensity level of profenofos solution is significantly lower than that of chlorpyrifos and methyl-parathion solutions of the same concentration, and its spectral contribution is the smallest, so the algorithm has poor resolution of profenofos in its mixed system. PARAFAC algorithm achieves the effect of “mathematical separation” instead of “chemical separation” that can qualitatively identify and quantitatively detect multi-component organophosphorus pesticide mixtures with serious spectral overlap without complicated preprocessing. The method provides a theoretical basis for rapidly detecting and analysing organophosphorus pesticide residues in water.