基于便携式近红外光谱仪的柴油硫质量分数检测  

Detection of Sulfur Content in Diesel Based on Portable Near Infrared Spectroscopy

作　　者：郑权 祝诗平[1] 齐宝华 唐鑫 黄华[1] ZHENG Quan;ZHU Shiping;QI Baohua;TANG Xin;HUANG Hua(College of Engineering and Technology,Southwest University,Chongqing 400715,China;Chongqing Hangtian Kaishi Co.Ltd,Chongqing 400718,China)

机构地区：[1]西南大学工程技术学院,重庆400715 [2]重庆航天楷世科技有限公司,重庆400718

出　　处：《西南大学学报（自然科学版）》2023年第3期232-238,共7页Journal of Southwest University(Natural Science Edition)

基　　金：重庆市自然科学基金项目(CSTC,2008BB1091)。

摘　　要：柴油中的硫是机动车排放造成大气污染的主要成分之一,目前各国政府制定了限定柴油硫质量分数的措施,因此研究对柴油中硫质量分数的快速检测方法具有重要意义.应用便携式近红外光谱仪采集不同硫质量分数柴油的光谱,共采集261份柴油的近红外光谱数据,利用Kennard-Stone(K-S)算法以3∶1比例将样本划分为校正集和预测集.对原始光谱在全谱区间采用去中心化、归一化、多元散射校正和15点2次平滑等多种预处理.实验结果表明,去中心化预处理方法对建立柴油硫质量分数的偏最小二乘回归(PLSR)模型效果最优,其决定系数(R2)为0.894和预测均方根误差(RMSEP)为89.17,相对分析误差(RPD)为3.089.比较了蒙特卡罗无信息变量消除(MCUVE)和竞争性自适应加权抽样(CARS)两种波长选择算法,最终使用CARS算法提取得到35个特征波长点进行高斯过程回归(GPR)建模的结果最佳,其R2为0.967,预测均方根误差为45.378,相对分析误差为5.616.结果表明,利用便携式近红外光谱技术建立柴油硫质量分数定量预测模型,实现对柴油中硫质量分数快速和无损的近红外定量检测具有可行性.Air pollution is a serious problem,of which the emissions of motor vehicles are their main pollution sources.Sulfur in diesel is one of the main components of atmospheric pollutants.At present,the measures to limit the sulfur content in diesel are formulated by governments all over the world.Therefore,the research on the rapid detection method of sulfur content in diesel is of great significance.A portable near infrared spectrometer(NIR spectrometer) was used to collect the spectra of diesel with different sulfur contents.A total of 261 diesel NIR data were obtained.Kennard stone(K-S) algorithm was used to divide the scale sample set to correction set and prediction set at a 3∶1 ratio.The original spectra were processed by various pretreatment methods such as decentralization,normalization,multivariate scattering correction,and 15 point twice smoothing.The experimental results showed that decentralized pretreatment method was proved to be the best for establishing the partial least squares regression(PLSR) model of diesel sulfur content,with the determination coefficient(R~2) of 0.894,the root mean square error of prediction(RMSEP) of 89.17,and the relative analysis error(RPD) of 3.089.Monte Carlo uninformative variable elimination(MCUVE) and competitive adaptive weighted sampling(CARS) were compared.Finally,Gaussian process regression(GPR) modeling with 36 characteristic wavelength points extracted by cars algorithm were the best,with R2 of 0.967,RMSEP of 45.378 and RPD of 5.616.The results show that it is feasible to establish the quantitative prediction model of sulfur content of diesel by using portable near infrared spectroscopy technology,and realize the rapid and nondestructive near infrared quantitative detection of sulfur content in diesel.

关 键 词：近红外光谱 柴油硫质量分数 便携式 波长选择 高斯过程回归 

分 类 号：O433.1[机械工程—光学工程] U473.12[理学—光学]