Large Eddy Simulation of Air Flow Around the Air-Sea Interaction Tower
Sponsored by the Office of Naval Research
Abstract
The dynamics of the coupled marine boundary layer (MBL) are driven by a myriad of interactive, multiscale and multiphase processes as described in “Coupled Marine Boundary Layer and Air Sea Interaction Initiative”. A thorough understanding of these processes would certainly have a profound impact on Naval worldwide operations. In order to study the MBL under the low-wind condition as part of the Coupled Boundary Layers and Air-Sea Transfer (CBLAST) program, the Air-Sea Interaction Tower (ASIT) has been constructed off the south shore of Martha’s Vineyard. The main objective of the work is to investigate the effect of flow distortion caused by the ASIT on the measurement data.
Results
In order to examine the flow around ASIT, it is necessary to fully understand the flow around a single wall-mounted cylinder. According to Kawamura et al. (1984), when the height-to-diameter ratio H/D is high enough (H/D>6), the periodic vortex shedding is expected along with the trailing vortex due to the end effect and the necklace vortex at a junction. As H/D is reduced, the periodic vortex shedding is suppressed in the range of H/D<2 (Okamoto and Yagita, 1973; Kappler, 2002). The flow around ASIT belongs to the long cylinder case with high H/D ratio (H/D~10).
In this research, we present large eddy simulation (LES) of flow around a wall-mounted circular cylinder with varying heights (H/D=2.5, H/D=10), and then flow around ASIT. The characteristic-Galerkin, fractional four-step finite element method (FEM) by Lin et al. (2005) is used for this purpose. For the case of H/D=2.5, the present calculation predicts the mean drag coefficient CD of 0.71 which lies in the range between CD obtained from the LES Smagorinsky model (CD=0.88) and the LES dynamic model (CD=0.60) by Fröhlich and Rodi (2004) who used a grid of almost 20 times denser than the present case. Experimental values available from the literature for the case with H/D=2 are CD =0.78 (Kawamura et al., 1984) and CD =0.73 (Okamoto and Sunabashiri, 1992). The good agreement between our calculation and experimental data suggests the present model is very accurate (4th order spatial accurate for convection term) and has optimal grid flexibility as compared with other existing LES models. The model has also been fully parallelized for high-performance large-scale computing.
For the case with H/D=2.5, streamtracer analyses are performed in various planes in the computational domain. In Fig. 1(a), the streamtracers are initially located in (-7.5D,y,0) plane (upstream) and most of them cannot pass through the cylinder since they hit the stagnation points. It is interesting to note that the streamtracers located near the bottom wall tend to have downward velocity near the stagnation points toward the junction of the cylinder and the bottom wall. This trend results in the downdraft near the cylinder junction as shown in Fig. 1(b) in which the streamtracers originally start in (0,y,0.6D) plane. In Fig. 1(c) the streamtracers are initialized behind the cylinder at (0.6D,y,0) to reveal flow separation.
The flow past a circular cylinder of height H/D=10 is also examined in the same computational setup as the case with H/D=2.5. As shown in Fig. 2(a), the streamtracers merge toward the center of z/D=0 plane, which is not apparent in the case of H/D=2.5. Averaged vertical structures in Fig. 2(b) reveal that the tip vortices and the necklace vortices are much stronger than the kárman vortices generated between the tip and the junction. Unlike the short cylinder case, the downward motion of the streamtracers behind the cylinder junction is followed by the strong upward motion at x/D~2 and further downstream.
The flow around the ASIT tower is much more complicated. In Fig. 3, the side-view of streamtracers released in the z/D = 0 plane using the averaged velocities is presented. Flow is accelerated inside the ASIT structure and exhibits downdraft near each of the three legs. The effect of the platform on the overall flow structure is minimal because of its slender shape.
The LES of the flow around the FLIP is also performed. The preliminary result is shown in Fig. 4. A large wake region is formed behind the vessel.
 |
Figure 1: Streamtracers for the mean velocity on plane. (a) Streamtracers initially located at (-0.75D,y,0). (b) Streamtracers initially located at (0,y,0.6D) (side of cylinder). (c) Streamtracers initially located at (0.6D,y,0) (behind cylinder).
 |
Figure 2: Streamtracers for the mean velocity in z = 0 plane. (a) Streamtracers initially located at (0,y,0.6D) (side of cylinder). (b) Streamtracers initially located at ( 0.6D,y,0) (behind cylinder).
 |
Figure 3: Sideview of streamtracers using the averaged velocities in the z/D = 0 plane. Contours on the streamtracers represent <u>.
 |
Figure 4: Oblique viewof streamtracers using the averaged velocities in the z/D = 0 plane around the FLIP. Contours on the streamtracers represent <u>.
