Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
– 5.28 –<br />
are moved to define the surface position. These procedures clearly imply<br />
computational simplification and approximation. Another method of surface<br />
positioning, which is based on the application of kinematic boundary condition along<br />
the surface, is discussed in Sect. 5.8.<br />
� Downstream of the cylinder, an over-estimation of the computed water surface is<br />
observed. Here also, a rather steep surface profile prevails, in which the high velocity<br />
(low flow-depth) in the plane � = 90° is readapting to the lower velocity (higher<br />
flow-depth) in the wake of the cylinder. Again, the method of surface positioning has<br />
certainly a low accuracy. In the scoured-bed case (see the next section), a similar<br />
over-prediction occurs; another interpretation of this problem, which attempts to<br />
relate it to the production of the turbulent kinetic-energy, will be discussed (see Sect.<br />
5.6.3). One expects that the disagreement of the water surface should disappear at a<br />
distance further downstream away from the cylinder. Unfortunately, the limited<br />
measured data do not allow such control. In the scoured channel bed case, it will be<br />
seen that the measured water surface further downstream returns back to the one of<br />
the approach flow.<br />
5.5.4 Conclusions<br />
The numerical model (see Chapter 4), conclusively, has demonstrated its applicability<br />
and ability to simulate flow around a cylinder on a flat channel bed. Detailed comparison<br />
of the computed velocity fields to the measured ones (see Fig. 5.10) shows satisfactory<br />
agreements. The downward and reversed flows upstream of the cylinder and the rotating<br />
flow downstream of the cylinder can be reproduced by the model (see Fig. 5.11 and Fig.<br />
5.12). A qualitative agreement is also seen from the vorticity fields (see Fig. 5.13). The<br />
model, however, shows a rather disagreement of the water-surface profile, notably along<br />
the cylinder circumference and behind the cylinder. A good agreement, nevertheless, is<br />
seen along the upstream water surface, including the bow wave at the cylinder upstream<br />
face (see Fig. 5.14). The method of surface boundary positioning, in fact, becomes less<br />
accurate in regions of a strong pressure gradient.<br />
5.6 Simulation of flow around a cylinder in a scoured channel bed<br />
5.6.1 Computational domain<br />
The computational domain covers a 5-[m] channel-reach with the cylinder at 2.5 [m]<br />
from the upstream boundary (see Fig. 5.15). The origin of the coordinate system is<br />
defined at the center of the cylinder, at the original uneroded channel bed. The flow is<br />
assumed to be symmetrical about the channel centerline and, therefore, the computational<br />
domain represents half of the channel with a symmetry plane on the left (north)<br />
boundary, y = 0. The grid is formed by 109 � 43 � 22 cells in the x, y and z directions,<br />
respectively. There are 103,114 computational nodes, 16,062 boundary nodes, and 704<br />
dummy nodes. The grid was generated by the same method as in the flat-bed case, i.e. by