On the Computation of Turbulent Mixing Processes with Application to EGR in IC-engines

University dissertation from Stockholm : KTH Royal Institute of Technology

Abstract: This thesis deals with turbulent mixing processes occuring in internal combustion engines, when applying exhaust gas recirculation (EGR). EGR is a very efficient way to reduce emissions of nitrogen oxides (NOx) in internal combustion engines. Exhaust gases are recirculated and mixed with the intake air of the engine, thus reducing the oxygen concentration of the combustion gas and the maximum combustion tempera- ture. This temperature decrease results in a reduction of NOx emissions, since NOx is produced at high temperatures.The issue of NOx reduction is of high importance for current engine development (particularly for heavy-duty engines), since NOx is the main cause for smog formation and subject to increasingly stronger emission legislation. One of the practical problems when applying EGR is the non-uniformity of the mixture among and inside the cylinders deteriorating the engine and emission performance.The aim of this work is to develop and assess methods suited for the computation of turbulent mixing processes in engine conditions. More specifically, RANS and LES computations are considered. The flow structures responsible for the mixing are analyzed for two different T-junctions and a six-cylinder Scania engine-manifold. Shortcomings and advantages of the applied mixing models are explained.The main results are, that commonly applied scalar flux models for the RANS framework do not predict correct scalar flux directions. In stationary flow, the applied k-?-model in combination with a gradient-diffusion-model gives too small mixing rates as compared to LES and experiments. Furthermore, the LES computations of the T-junctions show, that Dean vortices occuring due to the curvature of the flow are broken up and dissipated only a few diameters downstream of the junction. The RANS computations do not predict this break-up, giving fundamentally different flow structures and mixing distributions. In pulsating flow, a resonance between the natural stabilities and the pulsation frequency is found by LES results, which could not be predicted by RANS.Computations of the flow in a Scania intake manifold with generic boundary con- ditions indicate, that inlet pulsations are important for the mixing process and that the smoothing effect of URANS is not adequate for accurate mixing computations. LES, on the other hand, is more promising, since it is able to capture the physics of pulsating flows much better.

  CLICK HERE TO DOWNLOAD THE WHOLE DISSERTATION. (in PDF format)