Performance comparison of simulation models and parameters for engine CFD applications.

Abstract

In this thesis, the performance of computational fluid dynamics (CFD) simulation models and parameters are systematically compared for an engine model based on the Imperial college experimental engine. A set of simulation models and parameters were chosen to be studied. For each model and parameter, a few different options were chosen and simulation cases created for each possible different combination of these options. A custom code was written for creating the cases out of all the combinations of model and parameter options.

The simulations for this thesis were performed with OpenFOAM, which is an opensource CFD software. The engine model has been created at Wärtsilä and used in-house meshing tools to create a moving mesh.

After the simulations had ran, the results were post-processed using a custom code written for this thesis. The post-processed results were directly compared to results from a high-fidelity direct numerical simulation (DNS) study based on the Imperial college case, which has great agreement with the experimental results. For each simulation, two metrics were calculated that measured the difference between simulation and DNS results. These metrics were used to assess the performance differences between cases with different model and parameter combinations, and thus the performance of the models and parameters themselves.

The models and parameters studied in this thesis were the turbulence model, velocity advection scheme, maximum Courant number, mesh resolution and the number and thickness of surface inflation layers of the mesh. These were divided into three individual studies.

A few of the turbulence models had clearly better performance than the others. Change in the velocity advection scheme, maximum Courant number and surface inflation parameters had only slight effects on performance. An increased mesh resolution generally lead to better performance.