Search Results

Standard CFD methods require a mesh that fits the boundaries of the computational domain. For a complex geometry the generation of such a grid is time-consuming and often requires modifications to the model geometry. This paper evaluates the Immersed Boundary (IB) approach which does not require a boundary-conforming mesh and thus would speed up the process of the grid generation. In the IB approach the CAD surfaces (in Stereo Lithography -STL- format) are used directly and this eliminates the surface meshing phase and also mitigates the process of the CAD cleanup. A volume mesh, consisting of regular, locally refined, hexahedrals is generated in the computational domain, including inside the body. The cells are then classified as fluid, solid and interface cells using a simple ray-tracing scheme. Interface cells, correspond to regions that are partially fluid and are intersected by the boundary surfaces.

The flowfield around a 60% scale stationary Formula 1 tire in contact with the ground in a closed wind tunnel was examined experimentally in order to assess the accuracy of different turbulence modeling techniques. The results of steady RANS and Large Eddy Simulation (LES) were compared with PIV data, which was obtained within the same project. The far wake structure behind the wheel was dominated by two strong counter-rotating vortices. The locations of the vortex cores, extracted from the LES and PIV data as well as computed using different RANS models, showed that the LES predictions are closest to the PIV vortex cores. All turbulence models were able to accurately predict the region of strong downward velocity between the vortex cores in the center-plane of the tire, but discrepancies arose when velocity profiles were compared close to the inboard and outboard edges of the tire.

A generalized re-normalization group (RNG) turbulence model based on the local "dimensionality" of the flow field is proposed. In this modeling approach the model coefficients C₁, C₂, and C₃ are all constructed as functions of flow strain rate. In order to further validate the proposed turbulence model, the generalized RNG closure model was applied to model the backward facing step flow (a classic test case for turbulence models). The results indicated that the modeling of C₂ in the generalized RNG closure model is reasonable, and furthermore, the predictions of the generalized RNG model were in better agreement with experimental data than the standard RNG turbulence model. As a second step, the performance of the generalized RNG closure was investigated for a complex engine flow.

An integrated numerical model has been developed for diesel engine computations based on the KIVA-II code. The model incorporates a modified RNG k-ε, turbulence model, a ‘wave’ breakup spray model, the Shell ignition model, the laminar-and-turbulent characteristic-time combustion model, a crevice flow model, a spray/wall impingement model that includes rebounding and breaking-up drops, and other improved submodels in the KIVA code. The model was validated and applied to model successfully different types of diesel engines under various operating conditions. These engines include a Caterpillar engine with different injection pressures at different injection timings, a small Tacom engine at different loads, and a Cummins engine modified by Sandia for optical experiments. Good levels of agreement in cylinder pressures and heat release rate data were obtained using the same computer model for all engine cases.

Combustion simulations utilizing the characteristic time combustion model have been performed for four DI diesel engines ranging in size from heavy-duty to large-bore designs. It has been found that the pre-factor to the turbulent characteristic time acts as a scaling parameter between the engines. This phenomenon is explained in terms of the non-equilibrium behavior of the turbulent time and length scales, as is encountered in the rapidly distorting, spray-induced flows of DI diesel engines. In fact, the equilibrium assumption between turbulence production and dissipation, which forms the basis for the employed k-ε-type turbulence models, does not hold in these situations. For such flows, the real turbulent dissipation time scale is locally proportional to the turbulent characteristic time scale which is determined by a typical eddy turnover time.