Assessment of a Numerical Methodology for Large Eddy Simulation of ICE Wall Bounded Non-Reactive Flows 2007-01-4145

The increasing of the overall engine performance requires the investigation of the unsteady engine phenomena affecting intake air flow and the air-fuel mixing process. The “standard” RANS methodology often doesn't allow one to achieve a qualitative and quantitative accurate prediction of these phenomena. The aim of this paper is to show the potential and the limits of LES numerical technique in the simulation of actual IC engine flows and to assess the influence of some basic parameters on the LES simulation results. The paper introduces the use of a merit parameter suggested by Pope for evaluating the quality of the LES solution. The CFD code used is Fluent v6.2 and two basic test cases have been simulated. The first one is the flow over a backward facing step in order to perform a preliminary parametric numerical analysis. A one-equation dynamic subgrid-scales turbulence model is used. Then, an adaptive Large Eddy Simulation is performed in order to achieve a quasi-complete Large Eddy Simulation according to the merit parameter. A more complex engine steady flow bench case has been also simulated. A quasi-complete Large Eddy Simulation using the dynamic one-equation model and a simulation with the WALE sgs model have been performed. The quality of the LES results were checked using the merit parameter when the dynamic one-equation model was used. The quality of the results can be rated as reasonably accurate and useful for engineering investigation. If large scale motion effects must be evaluated in engine flows, computational time can be dropped by using the WALE model. The latter allows one to reduce the near wall refinement maintaining a reasonably accuracy in reproducing such large scale flow structures.