Research output: Contribution to Journal/Magazine › Journal article › peer-review
<mark>Journal publication date</mark> | 08/2009 |
---|---|
<mark>Journal</mark> | Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering |
Issue number | D8 |
Volume | 223 |
Number of pages | 16 |
Pages (from-to) | 1033-1048 |
Publication Status | Published |
<mark>Original language</mark> | English |
This paper aims to assess the feasibility of large-eddy simulation (LES) as applied to a real engine configuration. Two cases of LES were performed together with a reference Reynolds-averaged Navier-Stokes (RANS) simulation to study their performances in the prediction of the fuel injection and mixing in a high-speed direct-injection diesel engine. In addition, sensitivity of the LES results with respect to mesh resolution was investigated, as LES requires a higher mesh density and the filtering process depends on the mesh size. The LES meant that detailed information could be obtained on the fuel spray development and turbulent flow field that was not possible to capture using the RANS modelling approach. It was found that the LES captures flow structures that have a pronounced impact on the prediction of the fuel injection and mixing. High-resolution LES predicted higher liquid and vapour penetration into the cylinder, which is in better agreement with the experimental data obtained at Delphi Diesel Systems than is the RANS prediction. It was also shown that using a finer mesh in LES yields improvements in turbulent flow field prediction, at the same time reducing the contribution of the modelled part of the turbulent kinetic energy. Overall, it was proved that LES is a superior alternative to the traditional RANS approach for turbulent two-phase flows encountered in diesel fuel injections where flow unsteadiness is prominent. Improved predictions obtained from an advanced LES can subsequently enhance the understanding of the fuel injection process and assist the engine experimentation and design.