REVIEW: Models for predicting viscosities of biodiesel fuels over extended ranges of temperature and pressure

FERREIRA, ABEL G.M.; CARMEN TALVERA-PRIETO, NIEVES M.; PORTUGAL, ANTÓNIO A.; MOREIRA, RUI J.

FUEL

ELSEVIER SCI LTD

Año: 2021 vol. 287

0016-2361

Fuel viscosity is an important property that has a significant effect in fuel injection, spray development and combustion in Compression Ignition (CI) engines. Current and future injector designs of diesel engines (such as rail injection systems) work at high pressures (>100 MPa), meaning that fuel viscosity increases substantially over the atmospheric values. The estimation of biodiesel (BD) viscosity based on the knowledge of its composition would be of great potential in the optimization of biodiesel production processes, particularly in the blending of raw materials and refined products. In this work, comprehensive data sets were chosen from literature regarding several BD classes, in order to establish new correlations and new predictive methods of viscosity. The proposed methodologies were validated using available viscosity data of BDs having different chemical compositions in wide ranges of temperature and pressure. The new methods developed at atmospheric pressure for predicting BD viscosity were found to have better predictive ability than those commonly used in literature. In particular, the models developed with the Lewis and Squires equation fitted to biodiesel feedstock (LSDB model) and the same equation using the predicted degree of unsaturation (DU) (LSDU1 model) presented a very good performance with average relative deviation (ARD) < 2.5%. The results of those models were comparable to the uncertainty of the experimental measurements reported in the literature. For high pressure, the methods developed in the literature for oils and (in particular) vegetable oils were adapted here for biodiesel. A new method presented in this work using Avramov equation (AVR model) was able to predict the viscosity of different BD classes with ARD < 3%, which is better than that (ARD < 5%) predicted by the only method reported so far in literature.