Development of chemometric multivariate calibration models for spectroscopic quality analysis of biodiesel blends

Abstract

The fact that the biodiesel is produced from renewable resources and environmentally friendly when compared to the fossil-based petroleum diesel, biodiesel has gained an increasing interest. It is mainly produced from a variety of different animal fat and vegetable oil combined with an alcohol in the presence of a homogeneous catalyst and the determination of the quality of the produced biodiesel is as important as its production. Industrial scale biodiesel production plants have been adopted the chromatographic analysis protocols some of which are standard reference methods proposed by official bodies of the governments and international organizations. However, analysis of multi component mixtures by chromatographic procedures can become time consuming and may require a lot of chemical consumption. For this reason, as an alternative, spectroscopic methods combined with chemometrics offer several advantages over classical chromatographic procedures in terms of time and chemical consumption. With the immense development of computer technology and reliable fast spectrometers, new chemometric methods have been developed and opened up a new era for processing of complex spectral data. In this study, laboratory scale produced biodiesel was mixed with methanol, commercial diesel and several different vegetable oils that are used to prepare biodiesels and then several different ternary mixture systems such as diesel-vegetable oil-biodiesel and methanol-vegetable oil-biodiesel were prepared and gas chromatographic analysis of these samples were performed. Then, near infrared (NIR) and mid infrared (FTIR) spectra of the same samples were collected and multivariate calibration models were constructed for each component for all the infrared spectroscopic techniques. Chemometric multivariate calibration models were proposed as genetic inverse least square (GILS) and artificial neural networks (ANN). The results indicate that determination of biodiesel blends quality with respect to chemometric modeling gives reasonable consequences when combined with infrared spectroscopic techniques.