Large Eddy Simulation (LES) is a promising technique for turbulent combustion modeling. By resolving smaller scales in the turbulent flow, LES provides spatial averaged results rather than the ensemble averaged results from Reynolds Averaged Navier-Stokes (RANS) approach. The importance of LES can be seen that both industry and academia put increasing efforts on combustion modeling work based on LES.
This thesis work started with the modeling of sub-grid scale (SGS) scalar dissipation rate in non-reacting internal combustion (IC) engines. The model was formulated based on the a priori test of a direct numerical simulation (DNS) of force isotropic turbulence. A similarity term and a variable constant depending on the SGS Reynolds number were proposed and used in this model. A posteriori tests were performed by using the high-resolution measurement of in-cylinder scalar mixing field, and a multi-cycle LES simulation of the same engine.
The aforementioned model was further used to evaluate a mixing time scale based on mixture fraction variance and scalar dissipation rate in non-reacting and reacting IC engine flows. A chemical time scale, kinetically based on the specific internal change rate and chemical equilibrium state internal energy, was proposed and evaluated in reacting engine flows. These two time scales were used to regulate the direct chemistry solver (DCS) results in a computation cell. As a result, a mixing controlled direct chemistry (MCDC) combustion model is formulated and successfully implemented in the engine simulation code. This model was then tested in various operating conditions from a low temperature combustion (LTC) engine and a conventional, diffusion-combustion type diesel engine. To further evaluate the mixing time scale impact on the non-premixed combustion, a highly turbulent flame, Sandia flame E, was used for testing the capability of MCDC model to capture local extinction and re-ignition phenomena.
In addition to the scalar mixing and combustion modeling work, efforts were dedicated to evaluate the capability and ‘best practices for application’ of current engine LES code (KIVA3VLES). Proper orthogonal Decomposition (POD) methods were used in analyzing the velocity field. An investigation into the physical significance of POD modes and prediction converging rates were performed and presented. The potential benefits of POD analysis are suggested.