1 Fluid Mechanics, Department of Mechanical Engineering, Technical University of Denmark2 Department of Mechanical Engineering, Technical University of Denmark3 Energy Engineering, Department of Mechanical Engineering, Technical University of Denmark
In low speed large two-stroke marine diesel engines, uniflow scavenging is used to remove the exhaust gases from the cylinder and fill the cylinder with fresh air charge for the next cycle. The swirl enhances the mixing of fuel with air and improves combustion efficiency. The thesis focuses on characterizing the confined swirling flow during the scavenging process. A simplified experimental model of an engine cylinder is developed. Smoke visualization results show that at fully open intake port there is a well-defined vortex core. The core size increases in a hollow conical shape along the flow downstream. As the port closes, the mixing of smoke particles in the core with surrounding regions is enhanced. The hollow conical smoke pattern disappears and resembles to a jet. Laser Doppler Anemometry measurements are conducted in the swirl generator and at the entrance to the test cylinder. The results show that the incylinder swirling flow has a precessing vortex core. The precession frequency is found to be linearly dependent on the volumetric flow rate at a given swirl number. The stereoscopic particle image velocimetry (SPIV) measurements are conducted for two sets of experiments. In the first experiment, the intake port is kept fully open and three different cylinder lengths are investigated. The results indicate that the incylinder flow is a concentrated vortex decaying downstream due to wall friction. The mean axial velocity has a wake-like profile. The radial velocity is very small compared to tangential and axial components. No reverse flow is observed in the vortex core. The initially confined vorticity in the vortex core region is distributed to outer regions along the flow. Turbulent kinetic energy is high in the vortex core and near wall regions. The incylinder flow is majorly governed by the flow conditions at the cylinder inlet and the increased length of cylinder provides further decay of the swirl. The profiles of velocity components remain the same for a given cross-sectional plane common in different cylinder lengths. The mean position of the vortex center is not aligned with the cylinder axis at all measuring position and follows a helical path along the cylinder length. For cylinder length of eight diameters, the mean vortex path does not complete one revolution and instead re-twists at one side of the cylinder axis. In the second SPIV experiment, the measurements are conducted to characterize the effect of piston position on the in-cylinder swirling flow. The piston is positioned to cover the cylinder intake port by 0%, 25%, 50% and 75%. For increasing port closures the tangential velocity profile changes to a forced vortex and the axial velocity changes correspondingly from a wakelike to a jet-like. This change, however, starts at cross-sectional planes close to cylinder outlet and moves to upstream positions. At 50% port closure, the mean axial velocity in the whole cylinder attains a jet like profile. The tangential velocity resembles more to a wall-jet than a forced vortex profile. With 75% port closure, the jet-like axial velocity profile at cross-sectional plane close to intake port changes back to wake-like at the adjacent crosssectional plane and downstream. This indicates a vortex breakdown like characteristic. The tangential velocity then has forced vortex profile throughout the cylinder. The non-dimensional profiles of velocity components have no significant variation with the variation in Reynolds number. Numerical simulations are conducted only for the fully open intake port case. The turbulence models include RNG k and Reynolds stress models. The simulation results, however, do not show satisfactory agreement with the experimental data. The models predicted a larger vortex core size with a reverse flow. The downstream decay in the swirl is predicted to be lower than observed from experimental results. However, there are some qualitative features like distribution of modeled Reynolds stress components that, to some extent, have reasonable agreements. The factors affecting the performance of the CFD models possibly lie both in the treatment of turbulence and the numerical aspects.