Book Description

Open channels often have a rough bed surface. A good understanding of the characteristics of water flow over the rough bed is important for many applications in hydraulic engineering. Examples include investigations of the re-suspension and transport of contaminated bed materials, channel erosion, and the effectiveness of dissipating excessive flow energy. In this thesis, the rough surface was created by placing transverse bars at a flat channel-bed in one case and cubes in the other case, and the velocity field of rough-surface flow was computed through large eddy simulations (LES). This thesis aims to evaluate suitable computing strategies, validate computational results using available experimental data, and explore the detailed structures of near-bed flow. On the basis of the bars centre-to-centre spacing, , relative to their vertical dimension, k, studies of the rough-surface flow classify the rough elements as k-type and d-type ribs. Previous experimental studies of the rough-surface flow have produced some measurements of flow velocities, provided assessments of the suitability of existing methods for determining the shear velocity, and comparisons of characteristics between flow over d-type ribs and that over smooth walls, as well as between flow over ribs and that over transverse rows of staggered cubes. The main findings were that similarities existed in the outer layer between flow over three-dimensional roughness and flow over smooth walls. Some previous numerical studies introduced a form-drag term in the momentum equation to indirectly allow for the effect of roughness on the flow. Other numerical studies used obstacles as bed roughness elements, and applied non-slip conditions at their surfaces. A significant limitation of the previous studies is the use of conditions of fully developed boundary layer flow with zero-pressure gradient. This would be very approximate for water flows in shallow open channels. The canonical turbulent boundary layers are not expected to be valid, because the boundary layer thickness is a significant fraction of the depth of flow. Further investigations of the flow characteristics are needed. The large eddy simulations reported in this thesis capture the flow characteristics in the near-bed region. They gave numerical predictions of the flow field over transverse bars in shallow open channels at a range of /k ratios. The predictions provide accurate horizontal turbulence intensities, and horizontal velocity. It seems that large eddy simulations produce accurate flow characteristics in the streamwise direction, but less accurate results in the vertical direction. A plausible reason is the side wall effects. It is challenging to realistically capture secondary flow that creates transverse vortices in the transverse direction on both sides of the channel. These vortices cause upward motions at the middle of the channel. The power spectrum density (PSD) curves for all the runs show characteristics of energy cascade that are consistent with the results reported in the literature (Rodi, 2017), in particular the energy distribution for the inertia subrange. It has been concluded that: 1) an increase in the n/k ratio leads to a decrease in the region of reverse flow within the cavity, and an increase in the shear stress on the upstream face of transverse bars; 2) in the case of flow over staggered cubes, they tend to restrict reverse flow, and cause the shear stress at the upstream edge of the cubes to increase; 3) it is suitable to apply cyclic boundary conditions in both the transverse and streamwise directions in simulations of fully developed flow.