Best Practice Advice AC4-01: Difference between revisions
No edit summary |
m (Dave.Ellacott moved page SilverP:Best Practice Advice AC4-01 to Best Practice Advice AC4-01) |
||
| (10 intermediate revisions by 2 users not shown) | |||
| Line 1: | Line 1: | ||
{{AC|front=AC 4-01|description=Description_AC4-01|testdata=Test Data_AC4-01|cfdsimulations=CFD Simulations_AC4-01|evaluation=Evaluation_AC4-01|qualityreview=Quality Review_AC4-01|bestpractice=Best Practice Advice_AC4-01|relatedUFRs=Related UFRs_AC4-01}} | |||
='''Wind environment around an airport terminal building'''= | ='''Wind environment around an airport terminal building'''= | ||
| Line 22: | Line 24: | ||
The Underlying Flow Regimes (UFRs) associated with this AC are: | The Underlying Flow Regimes (UFRs) associated with this AC are: | ||
[[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]] Flow over surface-mounted cube/rectangular obstacles | |||
[[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]] is particularly relevant to this AC, since the geometry of the building is roughly rectangular in shape. Since the building is immersed in an atmospheric boundary layer UFR3-17 is also particularly relevant, but at the time of writing this UFR was not available. UFR3-02 also focuses on the atmospheric boundary layer, but for the mesoscale, rather then the local scale relevant to this AC. | |||
| Line 46: | Line 48: | ||
The extent of the computational domain should be large enough to place all boundaries sufficiently far from the buildings and areas of interest. Cowan et al. (1997) suggests that the computational domain around a building should not be smaller than 5H upstream, 15H downstream, and 4H at either side of the building. | The extent of the computational domain should be large enough to place all boundaries sufficiently far from the buildings and areas of interest. Cowan et al. (1997) suggests that the computational domain around a building should not be smaller than 5H upstream, 15H downstream, and 4H at either side of the building. | ||
The flow in question may have been transient and un-symmetric, but this element was not studied in the wind tunnel. Modeling half the domain imposes symmetry on the flow, even for a 3D calculation. Also, full 3D steady-state simulations can diverge, or yield alternative asymmetric solutions. Unsteady vortex shedding can be important for these calculations and can affect the re-attachment lengths. Time-dependent RANS solution and LES solutions have yielded better results under these circumstances [ | The flow in question may have been transient and un-symmetric, but this element was not studied in the wind tunnel. Modeling half the domain imposes symmetry on the flow, even for a 3D calculation. Also, full 3D steady-state simulations can diverge, or yield alternative asymmetric solutions. Unsteady vortex shedding can be important for these calculations and can affect the re-attachment lengths. Time-dependent RANS solution and LES solutions have yielded better results under these circumstances [[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]]. | ||
At the top and sides of the domain symmetry boundaries can be used. Representing the atmospheric boundary layer requires careful specification of inlet conditions, and in the case of a k-ε simulation, consistent profiles of k and ε, as well as mean velocity must be applied. The inlet boundary layer profiles need to be in equilibrium with the wall boundary (i.e. not change appreciably with distance downstream at the ground) otherwise the solution may yield physically unrealistic results. Although further work is required to fully resolve this issue, the profiles suggested by Castro and Apsley (1997), and Richards and Hoxey (1992) can be recommended. A rough wall boundary for the ground and a smooth wall boundary for the building are recommended, and it is important to check how the law of the wall and the roughness is specified in the CFD code, e.g. in the code Fluent various roughness parameters are used in lieu of the aerodynamic roughness z0 (which is used to specify inlet boundary conditions, as recommended above). | At the top and sides of the domain symmetry boundaries can be used. Representing the atmospheric boundary layer requires careful specification of inlet conditions, and in the case of a k-ε simulation, consistent profiles of k and ε, as well as mean velocity must be applied. The inlet boundary layer profiles need to be in equilibrium with the wall boundary (i.e. not change appreciably with distance downstream at the ground) otherwise the solution may yield physically unrealistic results. Although further work is required to fully resolve this issue, the profiles suggested by Castro and Apsley (1997), and Richards and Hoxey (1992) can be recommended. A rough wall boundary for the ground and a smooth wall boundary for the building are recommended, and it is important to check how the law of the wall and the roughness is specified in the CFD code, e.g. in the code Fluent various roughness parameters are used in lieu of the aerodynamic roughness z0 (which is used to specify inlet boundary conditions, as recommended above). | ||
| Line 55: | Line 57: | ||
Use a second-order or higher order scheme; it is generally accepted that these are less susceptible to errors than a first order upwind scheme. | Use a second-order or higher order scheme; it is generally accepted that these are less susceptible to errors than a first order upwind scheme. | ||
Use a level of grid resolution consistent with the scale of the building and refine the mesh locally to capture individual geometrical features or critical areas in the flow, e.g. at areas of high shear, separation etc. Castro et al. (1999) recommend a refinement of the order of 1%H around the building. Elsewhere in the domain a gradual expansion ratio (less than 1.2) can be used. This advice is not based on firm evidence, since a mesh-refinement sensitivity study was not carried out as part of this AC or | Use a level of grid resolution consistent with the scale of the building and refine the mesh locally to capture individual geometrical features or critical areas in the flow, e.g. at areas of high shear, separation etc. Castro et al. (1999) recommend a refinement of the order of 1%H around the building. Elsewhere in the domain a gradual expansion ratio (less than 1.2) can be used. This advice is not based on firm evidence, since a mesh-refinement sensitivity study was not carried out as part of this AC or [[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]]. Further guidance on mesh design and refinement near smooth wall boundary conditions is given in ERCOFTAC BPG (Casey and Wintergerste, 2000), but considerable care needs to be exercised when applying rough wall boundary conditions. | ||
Ensure that the solution is well-converged. There is evidence that a lower than normal tolerance for the residuals is required to avoid spurious results at the outlet and changes in boundary layer profile with downstream distance. | Ensure that the solution is well-converged. There is evidence that a lower than normal tolerance for the residuals is required to avoid spurious results at the outlet and changes in boundary layer profile with downstream distance. | ||
| Line 62: | Line 64: | ||
'''Physical Modelling''' | '''Physical Modelling''' | ||
Unsteady RANS is considered the best practicable method for modelling flow around buildings, since it captures the effect of vortex shedding which is an important factor affecting re-attachment lengths [ | Unsteady RANS is considered the best practicable method for modelling flow around buildings, since it captures the effect of vortex shedding which is an important factor affecting re-attachment lengths [[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]]. Calculations using LES are promising, but these are currently used in academic research rather than in practical industrial CFD modeling. | ||
Steady RANS models cannot be relied on for accurate computations of flow around single, surface-mounted bluff bodies, according to | Steady RANS models cannot be relied on for accurate computations of flow around single, surface-mounted bluff bodies, according to [[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]]. Although the mean flow was correctly predicted in this AC, this is likely to have been fortuitous, since there was a significant over-generation of turbulence around the building. Steady RANS are even less successful for isolated bodies which lead to strongly periodic wakes (e.g. vortex shedding). | ||
Model modifications to prevent production of excess turbulence energy in regions of strong strain may be used instead of the standard k-ε. However, these are known to give flow field predictions that deviate from experiment, even though they can lead to rather better surface pressure predictions [ | Model modifications to prevent production of excess turbulence energy in regions of strong strain may be used instead of the standard k-ε. However, these are known to give flow field predictions that deviate from experiment, even though they can lead to rather better surface pressure predictions [[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]]. The length of the downwind separation region, for example, is likely to be over-predicted by at least 35%, with consequently large errors in local flow velocities in the near wake. According to [[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]] there is no distinct advantage in using higher-order (steady) RANS. | ||
The use of wall functions in adverse pressure gradients and re-circulation regions (which occur at the building stagnation point and in its wake) is problematic but currently unavoidable. | The use of wall functions in adverse pressure gradients and re-circulation regions (which occur at the building stagnation point and in its wake) is problematic but currently unavoidable. | ||
| Line 72: | Line 74: | ||
Ground wall boundary conditions are the most difficult to apply, but can be modelled using rough wall functions. | Ground wall boundary conditions are the most difficult to apply, but can be modelled using rough wall functions. | ||
Two-layer k-ε models can be used, whereby a one-equation model in the near-wall region is used instead of wall-functions. However, this requires a much finer grid at the wall, to allow computation all the way to the wall [ | Two-layer k-ε models can be used, whereby a one-equation model in the near-wall region is used instead of wall-functions. However, this requires a much finer grid at the wall, to allow computation all the way to the wall [[Flow_over_surface-mounted_cube/rectangular_obstacles|UFR 3-14]]. | ||
| Line 119: | Line 121: | ||
Site Design and Implementation: [[Atkins]] and [[UniS]] | Site Design and Implementation: [[Atkins]] and [[UniS]] | ||
{{AC|front=AC 4-01|description=Description_AC4-01|testdata=Test Data_AC4-01|cfdsimulations=CFD Simulations_AC4-01|evaluation=Evaluation_AC4-01|qualityreview=Quality Review_AC4-01|bestpractice=Best Practice Advice_AC4-01|relatedUFRs=Related UFRs_AC4-01}} | |||
Latest revision as of 16:36, 11 February 2017
Wind environment around an airport terminal building
Application Challenge 4-01 © copyright ERCOFTAC 2004
Best Practice Advice for the AC
Key Fluid Physics
This Application Challenge relates to the flow around a large terminal building roughly rectangular in plan, with an aerofoil-shaped roof. It is immersed in an atmospheric boundary layer, which is assumed to be neutrally stratified. The surrounding area is relatively open, with a few neighbouring buildings of large plan dimensions but of lower heights, located upstream of the main terminal building.
The Design and Assessment Parameters for this AC are the velocity deficit and turbulence increase in the wake of the building. The key physics of the flow affecting these DOAPs are flow stagnation at the front of the building, flow separation at the roof of the building, the turbulent wake downstream, and the re-attachment of the flow on the ground, as illustrated in Figure 1 below.
Figure 1 Key physics - Wind environment around an airport terminal building
The Underlying Flow Regimes (UFRs) associated with this AC are:
UFR 3-14 Flow over surface-mounted cube/rectangular obstacles
UFR 3-14 is particularly relevant to this AC, since the geometry of the building is roughly rectangular in shape. Since the building is immersed in an atmospheric boundary layer UFR3-17 is also particularly relevant, but at the time of writing this UFR was not available. UFR3-02 also focuses on the atmospheric boundary layer, but for the mesoscale, rather then the local scale relevant to this AC.
Application Uncertainties
The building geometry is complex and contains a large amount of small scale detail, including small scale sharp edges, protrusions and cavities. Deciding which neighbouring buildings should be included is an additional complication. Nonetheless, a simplification of the building and omission of the neighbouring buildings can be justified if these do not affect the key physics. Although it is difficult to predict the effect of simplifications and the omission of the neighbouring buildings it is likely that this will only affect the flow locally. The characteristics of the wake well downstream of the buildings are expected to be primarily influenced by the size of the terminal building, rather than the detail of its geometry, though protrusions on the aerofoil-shaped roof may have an important influence in determining the separation point on the roof. Omitting the neighbouring buildings is also thought to be a second-order effect, especially since these are situated upstream of the terminal building and their heights are relatively low compared to the main building (~0.3H), As a result, the DOAPs for this Application Challenge, i.e. the velocity defect and turbulence intensity at the runway, would probably be unaffected.
The wind tunnel model was internally hollow, representing the actual internal building layout. It had several openings at ground level that allowed a small amount of flow to leak through the building. Although this may have an appreciable influence on the pressure difference between the areas upstream and downstream of the building, and the distribution of the pressure on the building cladding, the effect of this on the flow well downstream of the building is likely to be second-order.
The effect of the tunnel walls and wind tunnel blockage is difficult to include. Wind tunnel blockage effects are deemed to be significant when the cross-section of the model buildings exceeds 4% of the tunnel cross-section (Snyder, 1981). In this case, the terminal building is only 2.5% of the tunnel cross-section and thus tunnel wall effects could be ignored. Symmetry (rather than wall) boundaries were placed at the top and sides of the computational domain. The choice of suitable dimensions for the domain box was driven by the need to place the boundaries a sufficiently long distance away from the building and the points of interest at the runway. Although this is a subjective choice, it was backed by recommendations arising from previous simulations of bluff bodies (e.g. Cowan et al., 1997)
Specifying appropriate atmospheric boundary flow conditions at the inlet, for the mean velocity, turbulence kinetic energy and dissipation is an important area of uncertainty which can have a large effect on the DOAPs. The levels of turbulence specified at the inlet will affect the turbulence levels downstream of the building and thus the re-attachment and recovery of the flow.
Computational Domain and Boundary Conditions
Three dimensional effects are important and thus a 2-D idealisation must not be used; this was demonstrated in CFD2.
The geometry of the terminal building and the surrounding buildings can be simplified, providing the key physics are unaffected. This does however require significant judgment by the user.
The extent of the computational domain should be large enough to place all boundaries sufficiently far from the buildings and areas of interest. Cowan et al. (1997) suggests that the computational domain around a building should not be smaller than 5H upstream, 15H downstream, and 4H at either side of the building.
The flow in question may have been transient and un-symmetric, but this element was not studied in the wind tunnel. Modeling half the domain imposes symmetry on the flow, even for a 3D calculation. Also, full 3D steady-state simulations can diverge, or yield alternative asymmetric solutions. Unsteady vortex shedding can be important for these calculations and can affect the re-attachment lengths. Time-dependent RANS solution and LES solutions have yielded better results under these circumstances UFR 3-14.
At the top and sides of the domain symmetry boundaries can be used. Representing the atmospheric boundary layer requires careful specification of inlet conditions, and in the case of a k-ε simulation, consistent profiles of k and ε, as well as mean velocity must be applied. The inlet boundary layer profiles need to be in equilibrium with the wall boundary (i.e. not change appreciably with distance downstream at the ground) otherwise the solution may yield physically unrealistic results. Although further work is required to fully resolve this issue, the profiles suggested by Castro and Apsley (1997), and Richards and Hoxey (1992) can be recommended. A rough wall boundary for the ground and a smooth wall boundary for the building are recommended, and it is important to check how the law of the wall and the roughness is specified in the CFD code, e.g. in the code Fluent various roughness parameters are used in lieu of the aerodynamic roughness z0 (which is used to specify inlet boundary conditions, as recommended above).
Discretisation and Grid Resolution
Use a second-order or higher order scheme; it is generally accepted that these are less susceptible to errors than a first order upwind scheme.
Use a level of grid resolution consistent with the scale of the building and refine the mesh locally to capture individual geometrical features or critical areas in the flow, e.g. at areas of high shear, separation etc. Castro et al. (1999) recommend a refinement of the order of 1%H around the building. Elsewhere in the domain a gradual expansion ratio (less than 1.2) can be used. This advice is not based on firm evidence, since a mesh-refinement sensitivity study was not carried out as part of this AC or UFR 3-14. Further guidance on mesh design and refinement near smooth wall boundary conditions is given in ERCOFTAC BPG (Casey and Wintergerste, 2000), but considerable care needs to be exercised when applying rough wall boundary conditions.
Ensure that the solution is well-converged. There is evidence that a lower than normal tolerance for the residuals is required to avoid spurious results at the outlet and changes in boundary layer profile with downstream distance.
Physical Modelling
Unsteady RANS is considered the best practicable method for modelling flow around buildings, since it captures the effect of vortex shedding which is an important factor affecting re-attachment lengths UFR 3-14. Calculations using LES are promising, but these are currently used in academic research rather than in practical industrial CFD modeling.
Steady RANS models cannot be relied on for accurate computations of flow around single, surface-mounted bluff bodies, according to UFR 3-14. Although the mean flow was correctly predicted in this AC, this is likely to have been fortuitous, since there was a significant over-generation of turbulence around the building. Steady RANS are even less successful for isolated bodies which lead to strongly periodic wakes (e.g. vortex shedding).
Model modifications to prevent production of excess turbulence energy in regions of strong strain may be used instead of the standard k-ε. However, these are known to give flow field predictions that deviate from experiment, even though they can lead to rather better surface pressure predictions UFR 3-14. The length of the downwind separation region, for example, is likely to be over-predicted by at least 35%, with consequently large errors in local flow velocities in the near wake. According to UFR 3-14 there is no distinct advantage in using higher-order (steady) RANS.
The use of wall functions in adverse pressure gradients and re-circulation regions (which occur at the building stagnation point and in its wake) is problematic but currently unavoidable.
Ground wall boundary conditions are the most difficult to apply, but can be modelled using rough wall functions.
Two-layer k-ε models can be used, whereby a one-equation model in the near-wall region is used instead of wall-functions. However, this requires a much finer grid at the wall, to allow computation all the way to the wall UFR 3-14.
Recommendations for Future Work
Further work is needed to address uncertainties relating to specifying appropriate inlet conditions for modeling the atmospheric boundary layer [UFR3-17]
Calculations of the AC could be performed using the latest available models such as the Spalart-Allmaras (1992), Menter SST (1994), or Durbin V2F (1995).
Calculations using LES are particularly promising. LES techniques have been shown to perform generally very much better than RANS, perhaps largely because resolution of the large-scale unsteady motions produces the correct overall, time-averaged flow.
More detailed wind tunnel data for this or other realistic buildings would be particularly useful for the evaluation of CFD applied to real, practical wind engineering problems.
In this Application Challenge the range of DOAPs addressed was restricted to only two parameters due to lack of appropriate or good quality wind tunnel data. As a result, this Application Challenge focused on the unusual problem of wind conditions over a runway, and did not address more typical building-related wind engineering problems such as loads on the building cladding (mean and peak pressures), the ground level wind environment in and around the building and its effect on pedestrian wind comfort, and the interaction of the internal ventilation with the external flow through the building openings. Gathering additional data for CFD validation is strongly recommended.
References
Casey M. and Wintergerste T. (Ed.) (2000) “Best Practice Guidelines” European Research Community On Flow, Turbulence, And Combustion (ERCOFTAC), Version 1.0.
Castro I. P. (1979) “Relaxing wakes behind surface-mounted obstacles in rough wall boundary layers” Journal of Fluid Mechanics, Vol. 93, Part 4, pp.631-659.
Castro I. P. and Apsley D. D. (1997) Flow and dispersion over topography: a comparison between numerical and laboratory data for two dimensional flows” Atmospheric Environment, Vol. 31, No. 6, pp. 839-850.
Castro I. P. Cowan I. R. and Robins A.G. (1999) “ Simulations of flow and dispersion around buildings” Journal of Aerospace Engineering, pp. 145-160.
Cowan I. R., Castro I. P., Apsley D. D. (1997) “Numerical considerations for simulations of flow and dispersion around buildings” Journal of Wind Engineering and Industrial aerodynamics, Vol. 67&68, pp. 535-545.
Durbin P.A. (1995) Separated Flow Computations with the k-ε-v2 Model, AIAA Journal, V33, N4, PP659-664
Menter F.R. (1994) Two-equation eddy-viscosity turbulence models for engineering applications, AIAA J., vol.32, no.8,
Richards P. J. and Hoxey R. P. (1992) Appropriate boundary conditions for computational wind engineering models using k-ε turbulence model. J. Wind Eng. And Industrial Aero., vol. 52, pp. 394-399.
Snyder W. H. (1981) “Guidelines for fluid modeling of atmospheric diffusion” USEPA Report EPA-600/8-81-009, EPA, Office of Air Quality Planning and Standards, Research Triangle Park, NC.
Spalart, P.R. and Allmaras, S.R. (1992)"A One-Equation Turbulence Model for Aerodynamics Flows," AIAA Paper 92-0439
© copyright ERCOFTAC 2004
Contributors: Athena Scaperdas; Steve Gilham - Atkins
Site Design and Implementation: Atkins and UniS