This thesis presents the motivation, objectives and reasoning behind the undertaken PhDto investigate the capability of compressible Implicit Large Eddy Simulation (ILES) insimulating wall-bounded inhomogeneous flows with particular interest in the near wallregion and further presents the progress achieved to date. Investigation includes the assessmentof current ILES methods to resolve inhomogeneous turbulence as well as compressibleturbulent boundary layers and to improve on those models further.A channel flow is an excellent problem to use to investigate the properties of a SGSmodel near a wall. The presence of a solid boundary tends to alter the behaviour ofthe turbulent flow in a number of ways that need to be modeled by the SGS model inorder to correctly represent the flow near the wall and most importantly the boundarylayer. The presence of the wall inhibits the growth of the small scales, alters the exchangemechanisms between the resolved and unresolved scales and finally gives rise in the SGSnear wall region to important Reynolds-stress producing events.A literature survey was carried out to identify other numerical investigations in simulatingchannel flow as well as data that could be used for validation purposes. The mainparameters used to validate the level of resolution in simulating channel flow are identifiedand a number of tools are developed. The primary parameters extensively used to validateLES simulations of channel flow throughout the literature are mean flow velocity profiles,turbulent kinetic energy, dissipation and shear stress profiles, wall shear stress and frictionvelocities as well as energy spectra in the spanwise and streamwise homogeneousdirections.Compressible viscous ILES of inhomogeneous anisotropic turbulence in an incompressiblechannel flow at wall normal grid resolutions of 68, 96 and 128 cells are carriedout with grid clustering applied to the wall normal direction. Initial results conductedin the compressible regime show that in order to obtain satisfactory results, medium and fine grids are required whereas on coarser grids, some additional numerical method isrequired. Each reconstruction scheme introduces a numerical dissipation characteristic toitself that maybe regarded as a sort of turbulence model. Thus depending on the requireddissipation, a suitable limiter can be chosen.The investigation then moves on to supersonic turbulent flow incorporating shockboundarylayer interaction. Only the slope-limiters that prove to simulate the flow in thefully developed turbulent channel best are favoured and then also utilised in the subsequentcompressible ramp simulations. The capabilities of modelling the shock boundarylayer interaction, mean turbulent profiles and shockwave angle are investigated and comparedagainst those obtained by DNS simulations. It is found that the grid at the inlet ofthe ramp plays a significant role, since it needs to be fine enough to maintain the turbulentin flow at an acceptable level before reaching the shock-boundary layer interaction zone.Further, very high-order numerical reconstructions were found to have difficulties in remainingstable in the high gradient regions of the flow when formulated in conservativeform and therefore solutions were not possible to obtain. Nonetheless, lower order reconstructionmethods run smoothly and the momentum profiles obtained, matched closelythose obtained by DNS.
展开▼
