多环芳烃(polycyclic aromatic hydrocarbon, PAHs)具有强致癌性, 极大威胁着人类身体健康。因此, 寻找一种高效、精确的多环芳烃浓度检测方法十分必要。采用FS920荧光光谱仪分析了苯并(k)荧蒽（BkF）、苯并(b)荧蒽（BbF）、苯并(a)芘（BaP）混合溶液的荧光光谱特性。发现在激发波长260~400 nm、发射波长300~500 nm范围内, 混合溶液的荧光光谱重叠严重。当混合物浓度配比不同时, 荧光特性也存在很大差异。针对光谱图不能直接反映混合物各组分浓度的特点, 将人工蜂群(ABC)算法优化的径向基函数(RBF)神经网络应用于浓度检测中, 对比分析普通RBF和ABC-RBF神经网络模型。结果表明, ABC-RBF神经网络模型预测误差相对较小, 训练到95次时, 均方差精度达到10-3。BkF、BbF和BaP的回收率平均值分别为99.20%、99.12%和99.23%, 证明此网络适用于检测多环芳烃溶液, 为检测多环芳烃浓度提供了一种快速、有效的新方法。

Polycyclic aromatic hydrocarbons(PAHs) are a kind of organic pollutant which widely distribute in the environment and whose carcinogenicity is a great threat to humans health. It is necessary to find an efficient and accurate method to detect the concentration of PAHs. By analyzing the fluorescence spectra of the mixed solution of BkF、BbF and BaP, we can see that the fluorescence spectra of the mixed solution overlap seriously within the excitation wavelength range of 260－400 nm and emission wavelength range of 300－500 nm, respectively. There are large difference in fluorescence characteristics for different mixture concentration ratio of the mixed solution. Because the spectra can not directly reflect the concentration of each component in the mixture, we apply radial basis function (RBF) neural network with artificial bee colony (ABC) algorithm to the concentration detection. By comparing RBF and ABC-RBF neural network, we can draw a conclusion that the prediction error of ABC-RBF neural network is relatively small, and the average recovery rate of BkF, BbF and BaP is 99.20%, 99.12% and 99.23%, respectively.