基于BP神经网络的日落黄溶液荧光光谱的研究

对食用合成色素日落黄的荧光光谱进行了研究,发现在310~400nm紫外光的激励下,日落黄溶液发出强荧光,峰值荧光强度随浓度的增加先增强后减弱,且荧光谱峰位置出现明显红移。经分析认为,日落黄溶液能产生荧光是因为分子中偶氮键将一个苯环和一个萘环连接在一起,形成大共轭结构,并且取代基-SO3Na与-OH处于萘环的对位,大大增强了日落黄分子的共轭程度,使其具有强的吸光功能,发出强荧光。另外,结合BP神经网络,通过训练好的网络对4种不同浓度的样本进行浓度预测,结果表明相对误差分别为4.269%,6.078%,4.977%和5.308%,相对标准偏差分别为0.448%,0.375%,0.419%和0.414%。实验表明,该方法具有训练速度快、预测结果准确度高等特点,有望成为一种对食用合成色素进行高效、痕量检测的有效方法。     

Identification of tartrazine and sunset yellow by fluorescence spectroscopy combined with radial basis function neural network

The fluorescence spectra of synthetic food dyes of sunset yellow and tartrazine are analyzed. The fluorescence peak wavelengths of sunset yellow and tartrazine are 576 and 569 nm, respectively, while the fluorescence spectra widths are 480-750 and 500 750 nm induced by ultraviolet light between 310-400 nm. The fluorescence spectra of sunset yellow overlap heavily with those of tartrazine, so it is difficult to distinguish them. Based on the principle of radial basis function neural network, a neural network is obtained from the training of the 14 groups of experimental data. The results show that the species of sunset yellow and tartrazine could be recognized accurately. This method has potential applications in other synthetic food dyes detection and food safety inspection.     

基于径向基神经网络和BP神经网络对日落黄荧光光谱的分析

对食用合成色素日落黄的荧光光谱进行研究,发现在最佳激发波长370nm紫外光的激励下,荧光峰值波位于576nm;经分析认为,日落黄溶液之所以能产生荧光是因为分子中偶氮键将一个苯环和一个萘环连接在一起,形成大共轭结构,并且取代基与—SO3Na与—OH处于萘环的对位,大大增强了日落黄分子的共轭程度,使其具有强的吸光功能,发出强荧光。另外,结合径向基神经网络和BP神经网络对未知样本进行浓度预测,结果精确,平均相对误差分别为3.51%和5.45%,RSD分别为1.83%和2.95%。该方法有望成为对食用合成色素进行高效检测的有效方法。