There has been a surge in the use of computational fluid dynamics (CFD) to predict the impact of wind on buildings and their surrounding areas. While CFD is well established in certain fields, such as industrial engineering, transportation or aeronautics, it is still in its infancy in architectural terms. But why are wind studies necessary?
As buildings become taller and more complex, it is becoming increasingly important for designers to understand the effect that local wind forces will have on a development, especially at ground level.
Typically, wind tunnel tests are undertaken at the planning stage, in order to demonstrate to local authorities that a development will not cause unacceptable wind speeds to occur at pedestrian level. Or they may be left until later in the design process, when data is required for the detailing, where they can help with the cladding and/or building services' design.Armed with an accurate picture of the wind environment, the architect or engineer can predict the impact on:
lenvironmental performance (HVAC systems, natural ventilation, smoke exhaust);
linnovative schemes such as alternative energy systems using wind turbines;
lpollutant dispersion (traffic, boiler flues and industrial processes).
Flow diagrams The two most commonly used tools for predicting the behaviour of wind around buildings are wind tunnel testing and CFD. The question is, which assessment method should be used, and when?
Wind tunnel testing is well established and has a proven record of producing reliable results. However, the design team will need to secure access to a specialist test facility where the work can be undertaken by skilled engineers and technicians.
CFD, in contrast, is widely available. Software packages can be bought off the shelf and the work undertaken in-house or outsourced to an external consultant. But despite being apparently easy to use, CFD requires specific specialist knowledge at the input stage as well as at the interpretative stage. Similarly, it is not uncommon for a study to be contracted-out to a CFD expert who, although familiar with the concepts and workings of the program, might lack specific expertise in wind engineering. Lack of expertise might lead to uncertainty in the production and interpretation of the results.
Computational wind engineering is generally performed using commercially available CFD programs - which today are remarkably flexible in terms of the problems that they can address - but special care should always be taken not to underestimate the complexity of the procedures involved - not just in obtaining results, but also in assessing them.
Wind engineering has its own characteristics and peculiarities that make it difficult to model virtually.
For example, it is difficult to represent the incoming wind characteristics correctly - in particular, the air turbulence, which is influenced by the type of terrain surrounding the buildings. Flow patterns, even around a simple block building, typically comprise:
la horseshoe-shaped vortex on the windward side of the building;
ljets of air at the corners of the building that may extend for a considerable distance downstream of the building;
lregions of 'flow separation'on the top and sides of the building where the flow pattern 'separates' from the building;
la highly turbulent wake region on the downstream side of the building, where wind speeds are relatively low due to the sheltering effect of the building.
Incorrect representations of these flow features will lead to erroneous predictions of wind speed and surface pressures around the building.
Another fine mesh There are two other important issues surrounding the quality of a CFD prediction: the treatment of turbulence and the resolution of the numerical mesh, which is a conceptual diagram of predictive velocities and pressures at given points. These are predictive because all CFD programs currently have to make approximations and assumptions in the treatment of turbulence due to the practical limitations on the number of data points - and so mesh resolution - that can be handled without resorting to very large and powerful computers.
The most commonly used models solve the time-averaged equations, making use of empirical and semiempirical turbulence sub-models to represent the variations that are 'lost' in the averaging process. But because the equations are averaged, it is not possible to analyse peak wind loads, peak pollutant concentrations or gustiness.
This makes CFD unsuitable for studying structural responses. It could be used to assess the impact on building ventilation systems and pedestrian comfort, but only if information about gustiness is not required. With respect to numerical mesh diagrams, an inappropriate choice can lead to poor predictions.
This is true particularly in wind engineering applications, where sufficient resolution around the building envelope is critical if complex flow patterns are to be captured.
Like CFD, wind tunnel predictions require expert handling. Here, an important issue is the appropriate use of the scaling laws that allow the results obtained at model scale to be realised at full scale. A physical wind tunnel model may not need the same level of detail as an architectural presentation model; typically, only features greater than about 0.5m need to be represented. However, it is quite possible to use an architectural model rather than building a new one if it is at an appropriate scale.
Furthermore, wind tunnel tests can be carried out as quickly as computational modelling (and sometimes quicker, if data for numerous wind directions is required).
It is important to be aware of the benefits and limitations of each tool, and their suitability for different applications. CFD can be successfully applied to predict internal flows and to assess thermal comfort and air quality, but some designers may require the confidence that can only be gained from full-scale mock-up testing of complex solutions. Wind tunnel testing is still probably the most appropriate tool for examining external flows around the building and its impact on structural safety, pedestrian comfort and HVAC performance. In many instances, the most accurate and cost-effective, long-term solution may be to use both together.
lFor more information contact Philippa Westbury, head of environmental wind engineering at the Building Research Establishment, on westburyp@bre. co. uk, or telephone 01923 664300.
FURTHER READING lBS 6399: Part 2: 1997 Code of practice for wind loads. This standard replaced BS 6399: Part 2: 1995, which in turn replaced CP3 Chapter V: Part 2:
1972. BS 6399 is only appropriate if the response of the structure can be considered to be static; structures with a dynamic response are not covered.
lBRE Digest 436 part 2, Wind loading on buildings: BS 6399-2:1997, N Cook and R S Narayanan,1999, £10.50.
Worked examples effective wind speeds for a site, and loads on a twostorey house.
lWind loading: a practical guide to BS 6399-2,1999, £40, Nicholas Cook, Thomas Telford.
References The Foundation for the Built Environment (FBE) is sponsoring the BRE to consider the current status of wind testing models and to establish the relative benefits and limitations of wind tunnel analysis and CFD modelling; and to assess quantitatively their respective accuracy.