Large Eddy Simulation Of In-Cylinder Turbulence For Internal Combustion Engines

The mixing, transportation, and combustion of working fluid with air and the formation of emissions in internal combustion engine are strongly influenced by three-dimensional, highly unsteady and transient turbulent flows, which are caused by air intake process. The turbulent flows in cylinder directly affects the charging efficiency, as well as the air conditions, heat transfer effect between inlet and cylinder wall, and then affects the combustion which is the most important process in cylinder. Large eddy simulation(LES) method have the capacity for capturing the small and transient structures in in-cylinder, and it is very necessary for us to deeply understand the effects of these structures on many physical and chemistry processes. So the LES method was used to simulate the in-cylinder flows in this thesis.The key to use FLUENT software studying the intake flow process and turbulent flows of in-cylinder is choosing a right model, which containing the right selection of boundary conditions, initial conditions, mesh type and density, SGS model etc. In this thesis, ICEM CFD software was used to meshing piston-cylinder assembly with different buffer area, and a large number of cycles were computed. Computed mean and root mean squire (rms) velocity are compared to experimental measurements. By analyzing the comparison results, the effects of the factors on LES simulation results were discussed.The study reveals that large buffer area can make the intake flow process more realistic, and then improve the calculation accuracy. The effect of initial cycle exists, it makes calculation accuracy gradually reduced. With the number of cycles increasing, the effect of cycle-to-cycle variations will reduce until reaching a negligible level. Mesh type and density have an important impact on the computation results, the computed velocity of structured mesh or greater mesh density show better agreement with measurements. Compared to RANS and DES, LES provides more complete information on in-cylinder flow structure. Among several SGS models, dynamic Smagorinsky model and Kinetic-Energy Transport model show better performance, but which one is more suitable is to be confirmed. In addition, when FLUENT parallel computing is used under LINUX environment, it should be aware of that the number of CPU is not proportional to speedup of computing, the speedup is increasing very small when the number of CPU is over5.