Engine models for alternative fuels are available, but few are comprehensive, well-validated models that include accurate physical property data as well as a detailed description of the fuel chemistry. In this work, a comprehensive biodiesel combustion model was created for use in multi-dimensional engine simulations, specifically the KIVA3v R2 code. The model incorporates realistic physical properties in a vaporization model developed for multi-component fuel sprays and applies an improved mechanism for biodiesel combustion chemistry.
A reduced mechanism was generated from the methyl decanoate (MD) and methyl-9-decenoate (MD9D) mechanism developed at Lawrence Livermore National Laboratory. It was combined with a multi-component mechanism to include n-heptane in the fuel chemistry. The biodiesel chemistry was represented using a combination of MD, MD9D and n-heptane, which varied for a given fuel source. The reduced mechanism, which contained 63 species, accurately predicted ignition delay times of the detailed mechanism over a range of engine-specific operating conditions.
Physical property data for the five methyl ester components of biodiesel were added to the KIVA library. Spray simulations were performed to ensure that the models adequately reproduce liquid penetration observed in biodiesel spray experiments. Fuel composition impacted liquid length as expected, with saturated species vaporizing more and penetrating less. Distillation curves were created to ensure the fuel vaporization process was comparable to available data.
Engine validation was performed against a low-speed, high-load, conventional combustion experiments and the model was able to predict the performance and NOx formation seen in the experiment. High-speed, low-load, low-temperature combustion conditions were also modeled, and the emissions (HC, CO, NOx) and fuel consumption were well-predicted for a sweep of injection timings. Finally, comparisons were made between the results of biodiesel composition (palm vs. soy) and fuel blends (neat vs. B20). The model effectively reproduced the trends observed in the experiments.