Wind environment around an airport terminal building
Application Challenge 4-01 © copyright ERCOFTAC 2004
Comparison of Test data and CFD
The results of the simulation CFD2 and its comparison to the wind tunnel data was discussed previously in the results sections for CFD2. The main conclusions emerging from the results of these simulations are summarized below.
In CFD2 it was possible to produce incident wind profiles in the CFD calculations to match the velocity profile in the wind tunnel, however the corresponding turbulence kinetic energy k values were significantly lower than those measured in the wind tunnel for all CFD profiles. A reason for this discrepancy may be a deficiency in the boundary layer in the wind tunnel, i.e. it is possible that the profile was not fully developed at the location of the building and was not in equilibrium with the tunnel ground roughness. In CFD2, two different sets of inlet profiles were considered: one based on the established method of Castro and Apsley (BL2), and another of profiles (BL1) calculated using a 2D empty tunnel simulation based on a plausible modification of the value of Cμ to match the wind tunnel value of k at the wall. The BL1 mean velocity profile was similar to the experimental, and its k profile values were closer to those measured than the corresponding BL2 profile; nonetheless, using BL1 as input caused subsequent calculations of the flow around the building to diverge. Therefore the incident wind profiles for a CFD simulation should be selected with care; a converged solution was obtained using the profiles by Castro and Apsley (BL2), though further work is recommended to fully resolve this issue of selecting appropriate and self-consistent profiles, to match conditions in an atmospheric boundary layer.
The 3D simulations in CFD2 gave mean velocity profiles in the wake of the building that were in overall agreement with experimental data at the location of the runway, for a incident wind normal to the building (180°). Thus, the predictions for the velocity deficit (DOAP1) were relatively close to experimental data and are thus considered satisfactory for the purpose of assessing the wind environment over the runway.
Although the predictions for the mean velocity are acceptable, the prediction of the turbulence kinetic energy is considered less successful. CFD2 showed clear evidence of a large increase in the turbulence kinetic energy in the wake of the building (by a factor of 5 compared to the empty tunnel) which is larger than the turbulence increase in the wind tunnel measurements. The inadequacy of the k-ε turbulence model with regards to excessive k production is well-documented; in areas of high streamwise strain rates which occur around bluff bodies the k-ε model is known to fail to model the turbulence dissipation adequately and therefore over-predicts the turbulence kinetic energy in those areas. The roof protrusion and the roof leading edge are such areas, so one would expect excessively high quantities of turbulence to be generated there. As these unphysically high levels of turbulence are transported downstream, then these can cause an increase in the mean velocity as momentum is transferred from these high k areas to parts of the boundary layer that have been retarded by the building.
It is therefore interesting that the mean velocity predictions for both CFD2 should be close to experimental data despite the strong over-prediction of the turbulence in the wake. As indicated in CFD2 however, the actual levels of turbulence simulated in the wake (rather than the turbulence increase) did match the experimental results, but this is likely to be the result of the much lower levels of turbulence prescribed at the inlet balancing the over-generation of turbulence downstream of the building. This coincidence may have contributed to the simulations reproducing the flow re-attachment and recovery correctly, leading to a correct prediction of the velocity deficit at the runway.
As the building is roughly symmetrical and is much longer than it is wide, it may appear that a 2D simulation could be used instead of a full 3D calculation. In CFD2 a 2D simulation was carried out to illustrate why such a simplification is inappropriate; in contrast to the 3D simulation, the 2D simulation failed to predict any velocity deficit at the runway, since the flow had re-attached and recovered too soon, well before reaching the runway.
© copyright ERCOFTAC 2004
Contributors: Athena Scaperdas; Steve Gilham - Atkins
Site Design and Implementation: Atkins and UniS