Investigation of the effects of fatty acid components on fuel properties in biodiesel production

It is known that having detailed information about the structure, quantity, variety, and physical properties relationships of fatty acids and methyl esters is crucial for the production of biodiesel fuel that can meet the demand and perform well in diesel engines under different climatic conditions. In this study, ten vegetable oils were selected considering the differences in fatty acid profiles, and biodiesel production was carried out from these oils. The fuel properties of the produced biodiesels were analyzed in the laboratory using relevant equipment, and compatible results were obtained according to the TS EN 14214 standards. Gas chromatography analysis was conducted for all oil samples used to investigate the effect of fatty acid profiles. Machine learning methods were used to develop a model that could provide the relationship. Additionally, feature selection analysis was performed to create an ideal dataset since machine learning works based on the logic of learning from the data. Filter methods and wrapper methods were employed for the selection of features, which are the fatty acids. By applying these methods to various regression techniques, it was confirmed that there is a relationship between the cetane number and CFPP (Cold Filter Plugging Point) with the fatty acids. It was found that the correlation between palmitoleic acid and cetane number was very low for all three filter methods. The three most effective fatty acids for cetane number prediction were determined to be linoleic acid, linolenic acid, and oleic acid, while the effective ones for CFPP prediction were identified as palmitic acid, stearic acid, and erucic acid. Furthermore, it was concluded that oleic acid, linoleic acid, and linolenic acid have a negative effect. It was also determined that viscosity, pour point, and cloud point were not suitable for modeling.