Determination and quantification of kerosene in gasoline by mid-infrared and Raman spectroscopy

As a locomotive fuel, gasoline has become one of the main consumer products of Chinese car owners. Effective identification of gasoline adulteration is of great significance to personal safety, car maintenance and environmental protection. This study used mid-infrared and Raman spectroscopy to quantitatively analyze kerosene adulteration in gasoline. The content of kerosene in gasoline by mid-infrared and Raman spectroscopy was established by partial least squares (PLS), extreme learning machine (ELM) and random forest model prediction. Analyze the effects of Savitzky-Golay (S-G), Multiple scatter correction (MSC), standard normal variate (SNV), first-order, second-order and two pre-processing methods, and calculate the prediction results of the best pre-processing method under the model. The mid-infrared PLS model combined with SG superposition SNV preprocessing method, ELM model combined with SNV preprocessing method, random forest model combined with SG superposition 1st preprocessing method, the prediction correlation coefficients are 0.9828, 0.9374, 0.9817, respectively, And the root mean square error is 0.7878, 0.3606, 0.2175. The results show that the mid-infrared and Raman vibration spectroscopy methods can effectively detect the kerosene content in gasoline, while the mid-infrared PLS model combined with SG superposition SNV pretreatment method has the best effect The research is carried out for gas stations, production enterprises and supervision departments. (C) 2020 Elsevier B.V. All rights reserved.