第22卷第7期 2014年7月
文章编号
1004-924X(2014)07-1723-09
光学精密工程
Optics and Precision Engineering
Vol. 22No. 7
Jul.2014
近红外技术快速测定裂解焦油芳烃含量
刘逸，，张兆斌，张永刚，司宇辰，王国清
（中国石油化工股份有限公司北京化工研究院，北京100013）
摘要：针对常用的气相色谱(GC)分析法进行焦油芳烃检测分析周期较长的间题，本文用傅里叶变换型近红外光谱仪及相关软件对石脑油蒸汽裂解焦油芳烃的含量进行了测定和实验研究。考虑焦油样品颤色差异大、芳烃含量变化大，实验通过对样品管改进、异常值判断、建模波段及光谱预处理方法优选等优化了碳六、碳七、碳八、碳九、碳十及总芳烃含量6 组预测模型性能，建立了石脑油蒸汽裂解焦油样品中芳烃含量的快速分析方法，使得单个样品的分析时间缩短到2min 以内。优化后6组模型的相关系数（R)分别为0.99520，0.99308，0.94633,0.97899，0.94846，0,99863，交叉验证均方差(RMSECV)分别为1.07.0.806,2.17.0.979,0.665，1.15。未知样6组芳烃含量的近红外光谱及气相色谱（GC)分析数据吻合良好，t-检验绝对值均小于其临界t值(ta.05(17)2.11)。另外，近红外分析数据相对标准偏差（RSD)均小于 2%，显示提出的方法具有较好的重复性。
关键调：近红外光谱术；预测模型；芳烃；裂解焦油；含量测量
中国分类号：()657.33;TQ075
文献标识码：A
doi;10. 3788/OPE. 20142207, 1723
Determinationofaromaticsinnaphtha-crackedtar
by near infrared spectroscopy
LIU Yi',ZHANG Zhao-bin,ZHANG Yong-gang,SI Yu-chen,WANG Guo-qing(BeijingResearchInstitute of Chemical Industry,ChinaPetrochemicaland Chemical
Corporation(SinopecGroup)，Beijing1ooo13，China)*Correspondingauthor,E-mail:lyi.bihy@sinopec.com
Abstract: Gas Chromatography (GC) is usually used in analysis of aromatic contents of naphtha-cracked tar, however, it costs a longer analytical period. For solve this problem mentioned above, this paper applies Fourier Near-infrared Spectroscopy (NIRS) and corresponding software to the determination of aromatic contents of naphtha-cracked tar to shorten the analytical time and improve the analytical efficiency. Since cracked tars have a wide difference in both color and aromatic contents, modified sample tubes, outlier judgment, selective wave numbers and optimal preprocessing methods for NIR spectra were proposed to optimize 6 predict models of aromatics, by which the five models with carbon numbers from 6 to 10 and one for total aromatics were improved. Then, a rapid analysis method based on NIRS for determining aromatics in naphtha-cracked tar of stream cracking was established. With the method, the analytical time for per sample was shorten within 2 min. Experimental results show that the 6 optimized predict models can offer the correlation coefficients
收稿日期：2013-09-02；修订日期：2013-10-26.
基金项目：中国石化科研基金资助项目（No.G6001-11-11ZS0236）