Research On Detection Of Methanol Diesel Quality Based On Near-infrared, Mid-infrared And Raman Spectroscopy

With the rapid development of automobile industry and petroleum resources increasingly scarce, the alternative fuel for internal combustion engine oil products were to the attention of the countries. As a kind of important new alternative fuel for the diesel, The methanol diesel was widely development and research by the researchers around the world. However, uneven quality of the methanol diesel seriously affected its use in domestic promotion, so it is very important for the detection of methanol diesel quality. This article used methanol diesel as the object of experimental research on the near infrared, middle infrared and Raman spectroscopy to study and analysis the methanol content of methanol diesel and viscosity index for seeking a simple, fast, and environmental methanol diesel quality detection method and accurate and reliable quantitative detection model, which provided the basis for the popularization of methanol diesel in our country and the quality supervision. The main research results were as follows:1. With methanol diesel samples as the research object, using near infrared spectroscopy we quantitative analyzed methanol diesel methanol content and viscosity for methanol diesel samples, adopted and compared the effect of five kinds of pretreatment method: the smoothing, baseline correction, multiple scattering correction, normalization and the original spectrum etc., and the effect of different regression calibration method of the methanol content and the viscosity index were analyzed and compared. Results showed that: the prediction effect for the whole interaction validation Partial Least Square model of methanol content and viscosity of the multiple scattering correction was the most appropriate; Established PLS, Support Vector Machine and least squares support vector machines model of the methanol content, in the three kinds of mathematical regression method, the prediction effect of LS-SVM model was the best; For the viscosity, established the Principal Component Regression, PLS and LS-SVM three kinds of model, the prediction effect of LS-SVM model was the best.2. In this paper, Using mid infrared spectroscopy we quantitative analyzed methanol diesel methanol content and viscosity for methanol diesel samples, adopted and compared the effect of five kinds of pretreatment method: the smoothing, baseline correction, multiple scattering correction, normalization and the original spectrum etc., and analyzed and compared the prediction effect of PCR、PLS and LS-SVM model of the methanol content and the viscosity index. Results showed that: the prediction effect for the whole interaction validation Partial Least Square model of methanol content and viscosity of Baseline correction was the most appropriate; After dealing the spectral data with the baseline correction, the LS-SVM model prediction effect for methanol content and viscosity was the best. Among them, for the LS-SVM model of methanol content.3. In this paper, Using Raman spectroscopy we quantitative analyzed methanol diesel methanol content and viscosity for methanol diesel samples. Firstly, we adopted and compared the effect of five kinds of pretreatment method: the smoothing, baseline correction, multiple scattering correction, normalization and the original spectrum etc., Then, we selected the best pretreatment method to process the spectrum data as input, we used the continuous projection algorithm(SPA) for the screening of variables, and established PLS, SPA-PLS and SPA-LSSVM three kinds of prediction model, and compared the prediction effect of them. Results showed that: the prediction effect for the whole interaction validation PLS model of methanol content and viscosity of the MSC was the most appropriate; for methanol content and viscosity, the prediction effect of the MSC-SPA-PLS model was superior compared the MSC- PLS model, which illustrated the effectiveness of the SPA screening algorithm as variables extraction method; for methanol content and viscosity, the prediction effect of the MSC-SPA-LSSVM model had higher precision, and was the best.4. To sum up, the near infrared, mid infrared and Raman spectra three kinds of detection methods were all feasible for detecting methanol content and viscosity of methanol diesel, and had achieved very high detection accuracy, could satisfy real industrial testing standards. Among them, the detection effect of near-infrared spectroscopy was the best. In this study, the nonlinear LS-SVM models could get good prediction effect and higher prediction precision.