Diesel Fuel Spray Formation and Combustion - A Modeling Approach

Doctoral thesis, 2014

The formation, ignition and combustion of fuel sprays are highly complex processes, and the models that are traditionally used to describe them have a range of shortcomings. The introduction of commercial and non-commercial computational fluid dynamics (CFD) software has facilitated the development and application of new and improved multidimensional CFD models describing these phenomena. Spray modeling is generally done using the Eulerian-Lagrangian method, in which the gas phase is modeled as a continuum and the droplets are tracked in a Lagrangian way. Existing spray models vary widely in their accuracy and robustness, and the performance of spray simulations is strongly dependent on their discretization, as is the predicted level of spray-generated turbulence. The work presented here concerns the prediction of spray formation and combustion behaviors using improved models implemented in the free open source software package OpenFOAM. To enable improved RANS simulation of spray formation, the VSB2 spray model was implemented, refined, and tested under diverse ambient conditions. This model mainly differs from standard spray models in the way it describes the breakup and evaporation of the liquid fuel. It was designed to be unconditionally robust, have a minimal number of tuning parameters, and be implementable in any CFD software package that supports particle tracking. The results of VSB2 spray simulations were compared to data from the Engine Combustion Network (ECN) and experiments conducted at Chalmers University of Technology. In addition the Volume Reactor Fraction Model (VRFM), a novel partially stirred reactor model, was implemented in OpenFOAM. This model defines a reactor fraction based on the mixture fraction, a chemical progress variable, and their variances rather than defining mixing and chemical time scales. The effect of exhaust gas recirculation (EGR) on n-heptane sprays formed under Diesel engine-like conditions was studied using the VRFM and the results of these simulations were compared to experimental data from the ECN. Finally, results obtained using the VRFM were compared to data generated using the well stirred approach and the multiple RIF model.