Correlating the Cold Filter Plugging Point to Concentration and Melting Properties of Fatty Acid Methyl Ester (Biodiesel) Admixtures

Biodiesel is a renewable alternative diesel fuel made from plant oils, waste cooking greases, and animal fats. Its most common form is fatty acid methyl esters (FAME) from transesterification of lipids and methanol. Biodiesel has physical properties that compare well with conventional diesel fuel (petrodiesel). However, biodiesel has poor cold flow properties that must be monitored in cold weather. In this work, three correlation models are introduced that accurately calculate the cold filter plugging point (CFPP) of biodiesel. The models were developed using measured CFPP data from neat (unblended) biodiesel fuels and 24 binary biodiesel admixtures. The biodiesel fuels studied were from canola, palm, and soybean oils and yellow grease (CaME, PME, SME, and YGME). The solid-liquid equilibrium (SLE) model required accurate concentration, melting point (MP), and enthalpy of fusion (Delta H-fus) data for each FAME species in the mixture. These data were used to infer the SLE phase transition temperature (T-sLE) of the biodiesel mixtures. The T-sLE demonstrated a linear correlation (R-2 = 0.977) with CFPP. The MODified Empirical Correlation (MODEC) model (R-2 = 0.980) was obtained by analysis of (CFPP)(-1) versus ln(y(c16)) data where y(c16) is the mass fraction of methyl palmitate (MeC16). Finally, a second-order polynomial (R-2 = 0.982) was derived to calculate the CFPP from the modified long-chain saturation factor (LCSFmod), which was defined as a weighted average mass fraction of C16+ saturated-FAME (SFAME) in the mixtures. Weight factors were the MP of the corresponding SFAME species. Prevalidation tests on these models yielded good results for the calculated CFPP data. These performances exceeded those obtained by using 26 models from the scientific literature. The MODEC model performed better by a small margin than the other two new models. The main benefit of the MODEC model is that it requires just the y(c16) value instead of the complete compositional analyses needed to apply the SLE or LCSFmod models.