近红外光谱法检测乙醇柴油主要性能指标

乙醇柴油是柴油替代品的一种,它的使用越来越广泛,乙醇柴油品质由许多指标决定,采用传统方法检测这些指标不仅价格昂贵而且耗时长。近红外光谱技术是一种廉价、快速实时在线检测乙醇柴油品质的有效方法。本文采用近红外光谱技术结合最小二乘支持向量机检测了乙醇柴油的密度、粘度和乙醇含量,比较了线性和非线性校正技术(主成分回归、偏最小二乘回归和最小二乘支持向量机)对乙醇柴油品质的分析效果,同时也比较了不同预处理方法对预测模型能力的影响。实验结果表明,最小二乘支持向量机优于主成分回归和偏最小二乘回归模型,其对乙醇柴油密度、粘度、乙醇含量建模效果最优,相关系数分别是0.995 8、0.995 7和0.995 3;预测均方根误差分别为0.000 68、0.011 3和0.5714。

Detection of key performance indicators of ethanol diesel by the infrared spectroscopy method

Ethanol diesel is one of the alternatives for the diesel, which is used more and more widely. Many indicators reflect the quality of diesel ethanol fuel. It is not only expensive but also time consuming to detect these indicators with traditional method. Near infrared spectroscopy method is an inexpensive, fast and real time online test for the quality of ethanol diesel oil. In this paper, the density, viscosity and the quality of the ethanol content of diesel are detected, and the analysis effect on the quality of ethanol diesel are compared by linear and non-linear calibration technology, including principal component regression, partial least squares regression and least squares support vector machines(LSSVM), and the effects of different pretreatment methods on the prediction model capabilities are also compared. Experimental results show that LSSVM is better than the principal component regression and partial least squares regression model, with the optimal modeling effect on the density of ethanol-diesel, viscosity, alcohol content. The correlation coefficients are 0.995 8, 0.995 7 and 0.995 3, and the root mean square error of prediction are 0.000 68, 0.011 3 and 0.571 4, respectively.