摘要
采用人工神经网络方法和紫外分光光度法相结合研究多组分的同时测定,系统地研究了神经网络方法对苯酚、苯甲酸和苯胺混合样品同时测定的可行性。在大量实验数据的基础上采用matlab语言编写两层BP神经网络,并初步探讨了网络参数对网络预测性能的影响。结果表明,BP神经网络能在不改变样品性质的前提下实现了多组分的同时测定,最大限度地保护了样品不被破坏。方法的回收率为96.9%~109.3%,RSD为0.90%~1.28%(n=5)。通过对不同隐层神经元数目的研究发现,当隐层神经元数目达到100时,仿真结果和实验数据相似程度达到了90%以上,且随隐层神经元数目增多,训练误差下降速度加快,但训练速度变慢,训练时间延长。
The artificial neural network and ultraviolet spectrometry method were used for studying the simultaneous determination of multi - component. Simultaneous analysis of phenol, aniline and benzoic acid with artificial neural network based on experimental data and two BP layers of neurons by suing MATLAB program were studied. The effect of the number of the hidden layer of nerve cell on network forecast performance was discussed. The neural network not only realized the simultaneous determination of multiple components, but also obtained the maximum protection of sample so that the sample was not damaged. The recoveries were 96.9% - 109.3%, and the RSD were 0.90% - 1.28% ( n = 5 ). The effect of the number of hidden layer neurons was studied. The experiment results showed that the similarity of various simulation results and experimental data was more than 90% when the neuron ' s number of the hidden layer reached 100. When numbers of neuron incereased, training errors decreased quickly, but the speed of training became slowly, and the training time increased.
出处
《化学分析计量》
CAS
2008年第4期24-26,共3页
Chemical Analysis And Meterage
基金
安徽工程科技学院青年基金资助项目(2004YQ005)
关键词
人工神经网络
苯酚
苯胺
苯甲酸
紫外分光光度法
artificial neural network, phenol, aniline, benzoic acid, ultraviolet spectrometry method