论文标题:近红外光谱分析技术在油品分析中的应用研究
The Application of Near Infrared Spectroscopy in Oil Quality Analysis
论文作者 高俊
论文导师 姚成,论文学位 硕士,论文专业 应用化学
论文单位 南京工业大学,点击次数 173,论文页数 84页File Size848k
2005-05-01论文网 http://www.lw23.com/lunwen_680331597/ 近红外光谱; 偏最小二乘法; BP 人工神经网络法; 汽油; 润滑油基础油
Near-infrared spectroscopy;Partial least squares (PLS);Back propagation artificial neural network (BP-ANN);Gasoline;Lube base oil
在炼油过程中,通常需要对产品的关键品质比如汽油辛烷值、柴油十六烷值或润滑油粘度等进行在线监测。传统的实验室分析方法由于测定费用高,测量滞后大,而不适合实时在线分析。近红外(NIR)光谱分析技术是一种快速的无损分析技术,可用于对石油产品质量的在线实时分析。为此,本文对近红外光谱分析应用技术进行了比较深入的研究,具体包括以下几个内容: 1.通过大量的中外文献阅读,对近红外光谱分析技术及其在石油产品品质分析中的发展与应用研究做了一个较为系统完整的阐述,介绍了近红外光谱分析技术的原理及光谱预处理方法、多元校正建模方法和模型评价指标等技术以及近红外光谱技术在石油化工领域的应用现状。2.介绍了近红外光谱分析中的偏最小二乘(PLS)和反向传播人工神经网络(BP-ANN)等校正方法,根据实际应用需要,开发了一个BP-ANN定量校正应用软件,该软件应用于近红外光谱分析技术中的油品分析,对于处理非线性问题具有较好的分析效果。3.从炼油厂实际生产需要出发,选取分析频次较高的汽油样品85 个,用标准方法测定样品的辛烷值,采集其近红外光谱谱图,分别建立PLS 和BP-ANN 模型,并将这两种方法应用于汽油样品辛烷值的快速测定。研究表明,在1000～1800nm 波长范围内,近红外光谱技术用于预测汽油辛烷值中能以0.1 辛烷值单位的测量精密度和0.5 辛烷值单位的准确度预测出汽油辛烷值,所得的结果与标准的ASTM-CFR 辛烷值机测定的结果相一致。由于汽油的辛烷值与光谱信号之间的内在关系是非线性的,所以反向传播人工神经网络法(BP-ANN)所预测的结果要略好于偏最小二乘法(PLS)。与前人的研究相比较,BP 神经网络方法获得了更好的预测效果。4.在用高分辨毛细管气相色谱(GC)法测定成品汽油单体烃组成的基础上,应用偏最小二乘(PLS)校正方法,建立了近红外光谱法(NIR)快速测定汽油族组成(饱和烃、烯烃和芳烃)的分析模型。并将NIR 测定结果与高分辨毛细管气相色谱(GC)法分析汽油单体烃的测定结果进行了比较。实验结果显示,近红外光谱预测结果与标准方法测定结果的标准偏差(SEP)符合标准方法再现性要求,t检
Some properties of petroleum products in refinery such as gasoline octane number,diesel cetane number and viscosity of lube base oil are necessary to be on-line monitored. Laboratory analysis methods are usually not suitable for online monitoring because of their high cost and long time delay. Near-infrared (NIR) spectroscopy is a non-destructive real-time analytical method, and it is preferable to online monitoring for the properties of petroleum products. Therefore, this thesis researched the application techniques of NIR. The main contributions of this thesis are as follows: 1. Introduce the concepts and the principles of NIR quantitative analysis methods, multivariate calibration modeling and model evaluating index. Then review the applications and developments of NIR in petroleum products analysis. 2. Two chemometrics methods that back propagation artificial neural network (BP-ANN) and partial least squares (PLS) have been introduced. And one quantified correction applications of BP-ANN has been exploited. The software applies to petroleum products analysis in NIR and has better result for non-linear problem. 3. Based on 85 gasoline samples which motor octane number (MON) and research octane number (RON) were evaluated by ASTM-CFR engine in advance, two separate near infrared spectroscopy calibration models for gasoline octane number have been established by using BP-ANN and PLS. The mean standard deviations of the forecasted values were all less than 0.5 unit. For the relation between the gasoline octane number and the near-infrared absorbency is non-linear, experimental results show that the BP-ANN model gains higher prediction accuracy. Compared with former study, BP-ANN gains better prediction result. 4. Based on the results of high resolution capillary Gas Chromatography (GC) for a set of gasoline samples, the PLS calibration method is used to set up the calibration model of near infrared spectroscopy for determining major hydrocarbon classes content in gasoline. GC has tested the validity of the calibration models by comparison with results determined by the standard methods, which show that the accuracy of NIR method conforms with that required. Experimental results show that standard error of prediction (SEP) accord with demand of repeatability, the result of paired t test less than significance level 0.05 distributing value. Compared with GC, the NIR method has advantages of high speed, simplicity, better repeatability and lower analysis cost, etc. The NIR is suggested as a rapid method for the intermediate control analysis in gasoline production. 5. Applied Fourier transform-near infrared spectrometer, many kinds of lube base oils are determined in the range of near-infrared long wave. The PLS and BP-ANN calibration method are used to set up three calibration models of NIR spectra-viscosity index, 40℃viscosity and 100℃viscosity which are suitable for the lube base oils. Study result has shown that near infrared spectroscopy has capability to obtain the spectrum information that has the relationship with the viscosity of lube base oil. As a sort of non-linear method, back propagation artificial neural network (BP-ANN) is preferable to evaluate the relationship between the spectrum information and the viscosity of lube base oil.
|
