摘要
将同步荧光法结合人工神经网络(ANN)和支持向量回归(SVR)用于混合体系中荧光光谱重叠的菲(Phe)和2-乙基菲(2-EP)两组分的同时测定。通过三维同步荧光法结合平行因子(PARAFAC)分析寻得Phe和2-EP的特征波长差Δλ为118 nm。在220~280 nm范围内,以31个波长处荧光强度值作为模型的输入变量用于建立ANN和SVR模型。结果显示,ANN模型分析Phe和2-EP预测样本的回收率分别为92.5%~104.9%和96.1%~104.3%,预测均方根误差(RMSEP)分别为2.08和2.95;SVR模型分析预测样本的回收率分别为98.2%~101.3%和94.9%~104.2%,RMSEP分别为0.74和2.42。实际水样的加标回收实验显示,基体简单的矿泉水中两种模型均取得满意结果;而基体复杂的湖水中样品预测值较实际值低,且SVR模型比ANN的预测性能更加稳健,泛化能力更强。将同步荧光法结合SVR模型应用于Phe和2-EP单独及混合状态下与腐植酸(HA)相互作用的研究,结果显示混合体系中Phe和2-EP与HA的结合系数均小于各自单组分体系,表明Phe和2-EP之间存在竞争吸附。
A method was developed for the simultaneous determination of phenanthrene( Phe) and 2-ethylphenanthrene( 2-EP) in water by synchronous fluorescence spectroscopy combined with artificial neural network( ANN) and support vector regression( SVR). The characteristic wavelength offset( Δλ) is 118 nm by adopting three dimensional synchronous fluorescence spectroscopy with parallel factor analysis( PARAFAC). In range of 220-280 nm,the fluorescence intensities at 31 wavelengths were regarded as input variables for ANN and SVR models analysis. The results showed that the recoveries of Phe and 2-EP prediction samples were in the ranges of 92. 5%-104. 9% and96. 1%-104. 3%,respectively,obtained by ANN model,while the recoveries of Phe and 2-EP were 98. 2%-101. 3% and 94. 9%-104. 2%,respectively by utilizing SVR model. The prediction root mean square errors( RMSEP) of Phe and 2-EP were 2. 08 and 2. 95 with ANN model,meanwhile they were 0. 74 and 2. 42 with SVR model,respectively. Recovery experiments by standard addition in actual water samples demonstrated that a satisfactory result was obtained with the two models for mineral water( simple matrix). However,predictive values were lower than the actual values for lake water( complex matrix),in which the performance of SVR model was more steady,and had stronger generalization ability than ANN model. Moreover,the synchronous fluorescence spectroscopy coupled with SVR was used to simultaneously calculate the binding coefficients of Phe and 2-EP with humic acid( HA) in a mixture solution. The result showed that the bonding capabilities of Phe and 2-EP inthe mixed systems are both less than those in single component system,which indicated that there is a competing absorption existing between Phe and 2-EP.
出处
《分析测试学报》
CAS
CSCD
北大核心
2015年第12期1366-1371,共6页
Journal of Instrumental Analysis
基金
国家自然科学基金(21177102,21075102)
国家海洋局海洋溢油鉴别与损害评估技术重点实验室开放基金(201405)
关键词
同步荧光法
菲
2-乙基菲
人工神经网络(ANN)
支持向量回归(SVR)
synchronous fluorescence spectroscopy
phenanthrene(Phe)
2-ethylphenanthrene(2-EP)
artificial neural network(ANN)
support vector regression(SVR)