采用FS920稳态荧光光谱仪对绿茶和铁观音这两种不同品种茶叶的氯菊酯溶液的荧光光谱特性进行了分析,发现这两种茶叶的荧光峰均位于λex/λem=(390～410)/675 nm, 氯菊酯的荧光峰λex/λem=300/330 nm。为了准确测定这两种茶叶中氯菊酯农药残留的含量, 采用遗传算法优化的径向基函数神经网络对其进行了分析, 当训练到74次时, 均方差精度达到10-3, 绿茶、铁观音的氯菊酯溶液预测样本的平均回收率分别为99.35%和98.89%, 平均相对标准偏差分别为1.25%和1.21%。与建立的径向基函数神经网络模型进行了对比, 结果表明三维荧光分析技术与遗传算法优化的径向基函数神经网络相结合能够较好地检测出茶叶中氯菊酯农药残留的含量, 检测灵敏度大大提高, 检出限范围广, 可达0.004 8～24 mg/kg, 远低于欧盟规定的茶叶中氯菊酯最高残留限量0.1 mg/kg, 为检测农药残留提供了一种快速简便的新方法。

The fluorescence spectra of permethrin in green tea and tieguanyin were studied. The fluorescence characteristic peaks of green tea and tieguanyin existed at λex/λem=(390-410)/675 nm, while the permethrin fluorescence characteristic peak existed at λex/λem=300/330 nm. To determine the content of permethrin in teas, a new method of a radial basis function (RBF) neural network based on a genetic algorithm (GA) was proposed. When the training is to 74, the precision of the mean square deviation reaches 10-3. The average forecast recovery rates of permethrin in green tea and tieguanyin are 99.35% and 98.89%, the average relative standard deviation are 1.25% and 1.21%, and the detection limit range is from 0.004 8 to 24 mg/kg, which is far lower than that of the EU rules of permethrin in tea maximum residue limits standards. Through the contrast of the RBF neural network model, it is found that the three-dimensional fluorescence analysis technology combined with GA-RBF neural network can predict the content of permethrin pesticide residue in teas quickly and easily, the detection sensitivity and the precision are higher.