摘要
将新兴的多变量分析工具——独立分量分析(ICA)应用到石化行业中,分析柴油的十六烷值和密度、组分中芳烃含量与其近红外光谱之间的关系。ICA方法用于提取柴油近红外光谱数据矩阵的独立成分和相应的混合矩阵,再用BP神经网络对混合矩阵和十六烷值、芳烃含量以及密度进行回归分析,提出了新的柴油组分含量测定和特性分析的基于独立分量分析-神经网络回归(ICA-NNR)的近红外光谱分析方法。通过分析独立分量数和网络中间隐层的神经元数对模型性能的影响,分别建立柴油组分含量测定模型和密度特性的关联模型。结果表明,该方法用于实测的柴油样品近红外光谱数据的处理,测试样品集的标准方法测定值与所建模型预测值的相关性及相对误差均优于现行常用的PLS、PCR等方法。基于ICA-NNR的近红外光谱分析方法对石化行业中油品的组分及物理特性分析具有很好的可行性。
The emerging multi-variable analysis tool, independent component analysis（ICA） was introduced to the petrochemical industry applications, analyzing the content of cetane number and aromatics of diesel fuel and relations between its density and NIR. A new method of prediction of diesel content and characteristic analysis based on near-infrared spectroscopy, artificial neural networks and independent component analysis was proposed. In its application to diesel＇s near-infrared spectroscopy, the independent components and the mixing matrix were firstly extracted by the independent component analysis. And then, the regressions of the mixing matrix and cetane number, aromatics, density were done by the artificial neural networks. The influence of the numbers of independent components and the neurons in the hidden layer on the properties of model was further discussed, and three optimal analysis models were built respectively. This new chemometric method has been applied to the measured spectra of diesel fuel. The results of correlation coefficients and relative errors between the standard-method tested values and the near-infrared method predicted values are superior to the commonly used methods, such as PLS, PCR, etc. The results show the feasibility of establishing the models with ICA-NNR method for diesel＇s content and physical characteristics analysis in petrochemical industry.
出处
《石油学报（石油加工）》
EI
CAS
CSCD
北大核心
2008年第6期726-732,共7页
Acta Petrolei Sinica(Petroleum Processing Section)
基金
浙江省科技厅重点项目(2006C21044)资助
关键词
独立分量分析
人工神经网络
近红外光谱
柴油
independent component analysis（ICA）
artificial neural networks（ANN）
near infrared spectroscopy（NIR）
diesel fuel
