A relative assessment of chromatographic and spectroscopic based approaches to predict engine fuel properties of biodiesel

Abstract

Biodiesel produced from renewable sources is known to reduce carbon footprint compared to fossil diesel and thus, has a more significant potential for compression ignition engine applications. Unlike diesel, biodiesel can be produced from various sources and, thereby, subjected to composition variability. A priori knowledge of engine fuel properties of biodiesel would allow a careful feedstock choice for producing biodiesel for automotive and stationary engine applications. Regression models to predict engine fuel properties of biodiesel from its composition using Chromatographic and Spectroscopic-based approaches are explored in the present study to develop the most suitable method. Regression models are developed using composition and Fourier Transform Infrared spectra of seventy biodiesel samples. The composition of the seventy biodiesel samples measured using gas chromatography is correlated with the properties using multilinear regression. In the Spectroscopic-based approach, the seventy biodiesel samples' mid-infrared spectra are correlated with the properties using partial least square regression. The developed regression models based on Chromatographic and Spectroscopic approaches are validated using 33 external validation biodiesel samples. The results obtained show that the calorific value, cetane number, density, and kinematic viscosity of biodiesel samples are predicted with a mean absolute percentage error of 1.16%, 4%, 0.76%, and 2.6%, respectively, using the Chromatographic approach while it is 3.17%, 4.59%, 0.73%, and 3.66%, respectively with Spectroscopic method. The chromatographic approach, which resulted in better prediction than the Spectroscopic method, also captures biodiesel composition variations on the engine fuel properties.