应用PLS及BP神经网络定量分析汽油中甲缩醛含量  

Application of PLS and BP neural network method in quantitative analysis of methylal content in gasoline

作　　者：孙亮[1] 李丽华[1] 张金生[1] 

机构地区：[1]辽宁石油化工大学石油化工学院,辽宁抚顺113001

出　　处：《现代化工》2014年第6期146-147,149,共3页Modern Chemical Industry

摘　　要：建立了采用中红外方法测定汽油中甲缩醛含量的方法。通过实验室配比一定体积分数的甲缩醛-汽油混合物作为研究的基础数据,利用傅里叶红外光谱仪测定不同比例甲缩醛-汽油混合物的谱图,分别建立偏最小二乘法(PLS)和BP神经网络的中红外谱图分析的校正模型。其中偏最小二乘法甲缩醛定量模型的相关系数R2为0.975 3,预测均方根误差(RMSEP)为0.121;BP神经网络法甲缩醛定量模型的相关系数R2为0.974 2,预测均方根误差(RMSEP)为0.132。该方法是一种操作简便、快速可靠的分析甲缩醛含量的方法。A method for the determination of methylal content in gasoline products by mid-infrared spectroscopy is presented.Volume fractions of lab-prepared methylal-gasoline mixtures are used as basic experimental data.Spectra analysis of different proportions of methylal-gasoline mixture is determined by FTIR.BP neural network and infrared spectra analysis model of PLS are set up,respectively.For the quantitative model of PLS,the R2 and RMSEP are 0.975 3 and 0.121,respectively.For the model of the BP,the R2 and RMSEP are 0.974 2 and 0.132,respectively.It indicates that infrared method is a simple operated,rapid and reliable method for analysis of methanol content.

关 键 词：红外光谱 偏最小二乘法 BP神经网络 甲缩醛 主组分汽油 

分 类 号：TQ015.9[化学工程]