人工神经网络用于紫外光谱同时测定苯和甲苯及二甲苯的含量

SIMULTANEOUS UV-SPECTROPHOTOMETRIC DETERMINATION OF BENZENE TOLUENE AND XYLENE WITH APPLICATION OF ARTIFICIAL NEURAL NETWORK

根据反向传输（Backpropagation，BP）算法，应用三层ANN（人工神经网络，Artificial neural network）网络原理，对紫外光谱严重重叠的苯、甲苯和二甲苯的混合体系进行同时测定。在230～280nm范围内，以16个特征波长处的紫外吸光度作为网络特征参数，并通过均匀设计安排样本进行网络训练和计算。苯、甲苯和二甲苯的回收率依次为98．7％，99．4％和97．4％，测定结果的相对标准偏差分别为2．0％，2．8％和2．7％。

The problem of serious overlapping of the UV-spectra of benzene, toluene and xylene in their UV- spectrophotometric determination was solved satisfactorily by the application of the principle of three layers artificial neural network and the back-propagation algorithm, and simultaneous determination of the three homologous compounds was effectively performed. Sixteen absorbance values （A） at their characteristic wavelengths in the UV- spectral range of 230-280 nm were taken as network parameters, and the network training and calculation of the samples were carried out by the arrangement of uniform designing. Tests for recovery and precision were carried out by standard addition method at different concentration of the 3 homologues, giving values of average recovery of 98. 7%, 99. 4% and 97. 4%, and of RSD＇s of 2. 0%, 2. 8% and 2. 7% for benzene, toluene and xylene respectively.