Multidimensional Modeling of Natural Gas Jet and Mixture Formation in DI SI Engines: Development and Validation of a Virtual Injector Model

During the last years, the integration of computational CFD tools in the internal combustion (IC) engine design process has continuously been increased, allowing to save time and cost as the need of experimental prototypes has diminished. Numerical analyses of IC engine flows are rather complex from both the conceptual and operational sides. In fact, such flows involve a variety of unsteady phenomena, and the right balance between numerical solution accuracy and computational cost should be always reached. The present paper is focused on computational modeling of natural gas (NG) direct injection (DI) processes from a poppet-valve injector into a bowl-shaped combustion chamber. At high injection pressures, the efflux of gas from the injector and the mixture formation processes include compressible and turbulent flow features, such as rarefaction waves and shock formation, which are difficult to be accurately captured by the numerical simulation, particularly when combustion chamber geometry is complex and piston and intake/exhaust valve grids are moving. A three-dimensional moving grid model of the combustion engine chamber, originally developed by the authors, was enhanced by increasing the accuracy in the sonic section proximity of the critical valve seat nozzle, in order to precisely capture the expansion dynamics the methane undergoes inside the injector and immediately downstream from it. The enhanced numerical model was validated by comparing numerical results to Schlieren experimental images for nitrogen injection into a constant-volume bomb. Then, numerical studies were carried out in order to characterize the fuel jet properties and the evolution of mixture-formation for a centrally-mounted injector configuration in both cases of a pancake test chamber and the real-shaped engine chamber. Finally, the fluid properties computed by the model in the throat-section of the critical nozzle were taken as reference data for developing a new effective ‘virtual injector’ model, which allows the designer to remove the whole computational domain upstream from the sonic section of the nozzle, keeping the flow properties practically unchanged. The outcomes of such a virtual injector model were shown to be in very good agreement with the results of the enhanced complete injector model, confirming the reliability of the proposed novel approach.