基于拉曼光谱的乙醇柴油密度、粘度和乙醇含量分析研究  

Study on Density,Viscosity and Ethanol Content of Ethanol Diesel Based on Raman Spectroscopy

作　　者：欧阳爱国 张宇 唐天义 刘燕德 OUYANG Ai-guo;ZHANG Yu;TANG Tian-yi;LIU Yan-de(School of Mechatronics Engineering, East China Jiaotong University, Nanchang 330013, China)

机构地区：[1]华东交通大学机电与车辆工程学院,江西南昌330013

出　　处：《光谱学与光谱分析》2018年第6期1772-1778,共7页Spectroscopy and Spectral Analysis

基　　金：国家"十二五"(863)计划项目(SS2012AA101306);国家自然科学基金项目(61640417);江西省优势科技创新团队建设计划项目(20153BCB24002);南方山地果园智能化管理技术与装备协同创新中心(赣教高字[2014]60号);江西省研究生创新资金项目(YC2015-S238)资助

摘　　要：乙醇柴油作为清洁燃料是柴油很好的替代品,不同乙醇含量的乙醇柴油其粘度有差别,而乙醇的含量直接影响着柴油机燃烧性能。所以急需一种方法实现快速对乙醇柴油主要指标在线监测。对采集到的不同浓度的乙醇柴油的原始拉曼光谱数据使用Savitzkv-Golay平滑(S-G)、多元散射校正(MSC)、微分处理(1^(st )D和2^(nd)D)、标准正态变量校正(SNV)等四种方法以及他们的组合方法对光谱数据进行预处理后,分别建立了乙醇柴油密度、粘度和乙醇含量的偏最小二乘回归(PLSR)模型,比较不同的预处理方法发现,乙醇含量和粘度在S-G+2^(nd)D预处理后所建立的PLSR模型效果最好,预测集R_p分别为0.930和0.918,RM-SEP分别为1.237和0.034;S-G+1^(st )D预处理后所建立的乙醇柴油密度PLSR模型结果最优,预测集R_p最大,为0.962,RMSEP最小,为0.14×10^(-2)。将经过S-G+2^(nd)D预处理后的光谱数据选用递归偏最小二乘算法(RPLS)、无信息变量消除(UVE)、正自适应加权算法(CRES)、连续投影算法(SPA)四种变量筛选方法以及将它们组合筛选得到的波长变量分别作为输入变量建立了PLSR模型,在使用SPA-CARS波长筛选方法所建立的乙醇柴油乙醇含量的预测模型效果最优,其预测集的R_p,RMSEP分别为0.978 1和0.825 5。结果表明使用该方法可以很好的对乙醇柴油的密度、粘度以及乙醇含量等主要指标进行预测。Ethanol diesel as a clean fuel is a good substitute for diesel fuel.Different content of ethanol diesel oil viscosity is different,and the ethanol in the ethanol content of diesel fuel affects the quality of combustion,so we need have an instrument or method to achieve Ethanol Diesel Ingredients and Main Indicators Online Monitoring.In this paper,Savitzkv-Ethanol diesel as a clean fuel is a good substitute for diesel fuel. Different content of ethanol diesel oil viscosity is different, and the ethanol in the ethanol content of diesel fuel affects the quality of combustion, so we need have an instrument or method to achieve Ethanol Diesel Ingredients and Main Indicators Online Monitoring. In this paper, Savitzkv-Golay smoothing （SG）, multiple scattering correction （MSC）, differential processing （1stD and 2ndD）, standard normal variable correction （SNV）, and so on were used to calculate the original Raman spectra . The PLSR model of ethanol diesel density, viscosity and ethanol content was established by using different pretreatment methods for the original Raman spectra of different concentrations of ethanolic diesel. Compared with different pretreatment methods, the ethanol content and viscosity of PLSR model established by S-G＋2ndD pretreatment has the best effect, the prediction set Rp is 0.930 and 0.918 respectively, RMSEP is 1.237 and 0.034 respectively. The PLSR model of ethanol diesel density after SG＋1stD pretreatment is the best method. The optimal set of Rp is 0.962, RMSEP is 0.14×10-2. The wavelength model was selected by Recursive partial least squares（RPLS）, Uninformative variables elimination（UVE）, Compctitive Adaotive Reweighted Sampling（CRES）, Successive Projections Algorithm （SPA） wavelength screening method after S-G＋2ndD pretreatment, and the PLSR model was established. The prediction model of ethanol content of ethanol diesel oil was optimized by SPA-CARS wavelength screening method. The Rp and RMSEP were 0.978 1 and 0.825 5. The results show that the meth

关 键 词：拉曼光谱 乙醇柴油 偏最小二乘回归 波段筛选 

分 类 号：O657.3[理学—分析化学]