光谱自动检索算法在快速建立汽油光谱数据库中的应用.pdfVIP

石油学报(石油加工) 2017 年 1 月 ACTA PETROLEI SINICA CPETROLEUM PROCESSING SECTION) 第 33 卷第 1 期 文章编号: 1001-8719(2017)01-0131-07 光谱自动检索算法在快速建立汽油光谱 数据库中的应用 李敬岩，祷小立，陈 瀑，田松柏 (中国石化石油化工科学研究院，北京 100083) 摘要:首先采用聚类分析方法将某石油化工企业汽油样本进行适当分类，然后采用光谱库自动检索算法，从 RIPP 汽油库中有针对性地找到一定量的汽油样本作为校正集，建立汽油重要性质的分析模型。近红外分析方法结合偏 最小二乘法具有测量快速、操作简单、无需预处理、重复性好等优点。建立的校正模型对该石油化工企业汽油样 本的辛烧值、烯炬含量和芳短含量的预测标准偏差分别为 O. 3 、1. 6%和1. 0% ，满足快速分析要求。 关键词:汽油;偏最小二乘(PLS) ;近红外光谱(NIR) ;库检索z 聚类分析 中国分类号: 0657.33 文献标识码: A doi: 10. 3969/j. issn. 1001-8719. 2017. 0 1. 018 Applicatioo of Spectral Automatic Retrieval Algoritbm 00 tbe Rapid Establisbmeot of Gasolioe Spectral Database LI Jingyan , CHU Xiaoli , CHEN Pu , TIAN Songbai CResearch Institute of Petroleum Processing , SINOPEC , Beijing 100083 , China) Abstract: The petrochemical enterprise gasoline samples were properly classified by cluster analysis method , then the spectral library automatic retrieval algorithm was employed to find a certain amount of gasoline samples as calibration sets from RIPP gasoline library to establish the analysis model for the properties of gasoline. The calibration models of research octane number (RON) , olefin and aromatics content had been established by Partialleast square (PLS). The standard error of prediction CSEP) of the petrochemical enterprise gasoline samples on gasoline RON , olefin and aromatics mass fraction were O. 3 , 1. 6 % and 1. 0 %, respectively. The results predicted by this method were very close to those determined by stan