Flash point prediction: Reviewing empirical models for hydrocarbons, petroleum fraction, biodiesel, and blends

Flash Point (FP) of fuels is essential to determine the requirements for transportation and storage temperature. FP may be a restriction for incorporating light fractions into final products; therefore, accurate prediction models are necessary. However, finding the most accurate prediction model might be a challenge due to the high variety of methods. Hence, we reviewed empirical correlations, the simplest approach to predict FP of pure hydrocarbons, petroleum fractions, blends and biodiesel-FAME once it does not require complex calculation and known composition. Those correlations originally adjusted for pure hydrocarbons were extended to predict FP of petroleum fractions and, vice-versa, to verify the possibility of extension of the models. In addition, empirical models based on blending indexes were tested to predict FP of blends of diesel-biodiesel and hydrocarbons, which were also compared to Liaw's model. The empirical models were assessed using experimental data from literature and from Brazilian refineries. As a result, there is not only one reasonable method for FP estimation of those compounds. It depends on the type of hydrocarbons, range of FP, and ideality and volumetric fraction for blending systems. In addition, we obtained a reasonable result for biodiesel FP prediction from specific vegetable oils by applying kinematic viscosity with AAD = 4.14 K. For petroleum fractions (distillation range from 340 K to 505 K), unexpectedly, a method firstly developed for pure compounds presented the most satisfactory performance with AAD = 8.11 K. Finally, we developed a new model to estimate flash point of Brazilian diesel fractions S500 with AAD = 2.11 K.