Rapid Measurement Of Biodiesel Conversion Rete Based On Near-infrared Spectroscopy

This thesis aims to apply Near-infrared (NIR) spectroscopy analysis technique to detect the biodiesel conversion rate rapidly. Biodiesel is one of the best solutions to solve the environmental pollution and the shortage of fossil fuel in today’s society. Foreign countries especially the developed countries have made large investment in the field of scientific research, and have achieved substantial results so far. Currently, related research mainly focused on using gas chromatography (GC) or high performance liquid chromatography (HPLC) to analysis the content of biodiesel or physical parameters. Although NIR spectroscopy technology has been reported to monitor the biodiesel conversion rate, this analysis technique is not illustrated with a systematic way. On the contrary, domestic research of biodiesel conversion rate is sorely lacking and applying NIR spectroscopy technology to measure biodiesel conversion rate is much scarcer.The main contributions of this thesis are as follows:1. Introduce the related background of biodiesel systematically, present the research and application of biodiesel in foreign countries, and highlight its significance to our country. Secondly, point out the importance of chemometrics in NIR spectroscopy developments based on the deep insight of NIR spectroscopy principles, features and applications. Finally, establish the research system of applying NIR spectroscopy technology to monitor biodiesel process.2. Introduce the three aspects of spectral pretreatment, wavelength selection method, modeling of chemometrics in the field of NIR spectroscopy. It has been generally recognized that deleting uninformative wavelengths can improve prediction performance, the stability of calibration model, and decrease modeling time, therefore, this study tries to make a breakthrough in the theory of wavelength selection algorithm. Based on the intensive study of the moving window partial least squares algorithm (MWPLS), an objective wavelength selection method named interval moving window partial least squares (iMWPLS) is proposed mainly to overcome the possible subjectivity introduced by MWPLS. This new objective procedure is applied to two real standard NIR datasets. Results demonstrates that iMWPLS can achieve an effective wavelength selection and improve predictive accuracy in NIR spectroscopy.3. Obtain a reliable field biodiesel conversion rate data after building biodiesel process platform, collecting NIR spectra, measuring the real conversion rate by nuclear magnetic resonance spectroscopy (NMR), and excluding abnormal sample points. Then, through the comparison of multiple regression models, the partial least squares regression method is determined to be the best model; through the comparison of various pretreatment methods, the Savitzky-Golay smoothing method is determined to be the best pretreatment method; through the comparison of multiple wavelength selection methods, successive projection algorithm (SPA) is determined to be the best wavelength selection method. After these steps, the eventual regression model uses only18wavelength points, and the root mean square error (RMSEP) is0.0115, coefficient of determination (R2) is0.9877.4. The proposed method iMWPLS also achieves relatively good results in the biodiesel NIR spectroscopy. Compared with a variety of current popular wavelength selection algorithms, the performance of the PLS model based on iMWPLS algorithm is close to SPA and is significantly better than UVE, iPLS, SiPLS, BiPLS and MWPLS.