ABSTRACT
Recently, serious scientific and technological attention is paid to creation of alternative energy sources, including biofuels. The assessment of the quality of the biofuels produced and of the raw materials needed for the production technology is an important scientific challenge. One of the major sources for biodiesel production is plant oils material (sunflower, rapeseed, palm, soya etc.). Since plants are complex system from the biota it is not easy to find specific chemical components responsible for their ability to serve as biodiesels. The characterization and classification of plant sources as biofuel material could be reliably estimated only by the use of multivariate statistical approaches (chemometrics). The chemometric expertise makes it possible not only to classify different biofuel sources into similarity classes but also to predict the membership of unknown by origin chemically analyzed samples to already existing classes. The present study deals with the prediction of the class membership of several unknown by origin samples, which are included in a large data set with FAME profiles of biodiesel plant sources. Using a data set from chromatographic analysis of fatty acid methyl esters profiles (FAME) of different plant biodiesel sources and applying the chemometric technique know as partial least squares-discriminant analysis (PLS - DA) a pattern recognition procedure is developed to: I. Model classes of similarity of biodiesel plant sources using their FAME profiles not taking into account the samples with unknown origin; II. Classify correctly the samples with unknown origin to the previously defined classes of biodiesel sources (palm oil, soybean oil, peanut oil, rapeseed oil, sunflower oil and maize oil). The prediction is successfully achieved for all samples with previously unknown origin. This pattern recognition approach is applied for the first time in the field of biodiesel classification and modeling tasks.
