摘要
利用衰减全反射傅里叶变换红外光谱(ATR-FTIR)技术分别对毒死蜱、炔螨特的微量溶液进行了检测,采用差谱、基线校正和矢量归一化对谱图进行预处理,利用BP神经网络分别使用自适应调整学习率并附加动量因子的梯度下降反向传播算法训练函数和SCG反向传播算法训练函数建立了毒死蜱和炔螨特农药溶液的定量分析模型,并对校正集和预测集进行了定量分析.毒死蜱溶液模型的分析结果为:R=0.998 6,RMSEC=0.100 0,RMSEP=0.220 1;炔螨特溶液模型的分析结果为:R=0.997 4,RMSEC=0.391 8,RMSEP=0.624 1.结果表明,BP神经网络结合ATR-FT-IR技术检测微量农药溶液含量具有快速、精度高、泛化能力强的优点,可用于农药溶液含量的快速、准确鉴定.
ATR-FTIR technology was used in this article to detect trace element solution of Chlorpyrifos and Propargite separately. Subtraction, baseline correction and vector normalization was used to prepro- cess the spectrum data. The quantitative analysis models of Chlorpyrifos and Propargite solution were established using the function traingdx and trainscg of MATLAB BP Neural Network toolbox. The experiment results showed that for the chlorpyrifos determination, the correlation coefficient was 0. 998 6, root mean square error of cross validation was 0. 100 0, and root mean square error of prediction was 0. 220 1 ; for the propargite determination, the correlatin coefficient was 0. 997 4, root mean square error of cross validation was 0. 391 8, and root mean square error of prediction was 0. 624 1. As results indicated, application of BP Neural Network in detecting trace pesticide solution based on ATR-FTIR Technology was a quick and precisely method with good generalization ability.
出处
《北京工商大学学报(自然科学版)》
CAS
2011年第4期64-67,共4页
Journal of Beijing Technology and Business University:Natural Science Edition
基金
北京市自然科学基金项目(4073031)
北京市优秀人才资助项目(20081D0500300130)