Computational fluid dynamics for buildings

[edit] Introduction

Computational fluid dynamics (CFD) is a technique used to model the behaviour of fluids.

In building design it is typically used to model the movement and temperature of air within spaces. This is important as it allows designers to investigate internal conditions before a building is built, allowing them to test options and select the most effective solutions.

CFD can be used for modelling:

The thermal comfort of occupants.
The distribution of environmental conditions within a space.
The effectiveness of building services (such as the positioning of air inlets and extracts or radiators).
The consequences of fire (such as the spread of heat and smoke).
The effectiveness of natural ventilation (such as the stack effect).
The build up of heat in specialist spaces such as server rooms.
The positioning of sensors. For example in a tall space, the temperature at the top might be very different to the temperature at the bottom. This can be important when positioning temperature sensors that feed back to the building management system. Otherwise, heating and cooling might be operating unnecessarily.

CFD can also be used to investigate the impact of a new building on air movement around a site, and has been used to model other 'fluid' behaviour, such as the movement of people.

Simulations are typically run for a number of different scenarios, testing behaviour under different levels of occupancy, different climatic conditions, in different modes of building services operation, with different openings between spaces and so on. This can build up an overall picture of how a building is likely to behave under normal operating conditions as well as during unusual or extreme conditions.

[edit] How it works

CFD works by dividing a space into a grid containing a large number of 'cells'. The grid of cells is surrounded by boundaries that simulate the surfaces and openings that enclose the space. The temperature of the boundaries, the air movement at openings, and the air temperature within the cells is then set to a starting condition which it is hoped is close to those that might be expected to be found within the space. These conditions might be determined using a boundary model that predicts boundary conditions, based on climatic and materials data. The more accurate the boundary model, and the closer the starting condition is to the final position predicted by the model, the faster the model will run and the more accurate the output is likely to be.

The software will then simulate the flow of air from each cell to those surrounding it, and the exchange of heat between the boundary surfaces and the cells adjacent to them. After a series of iterations, the model will come to a steady state that represents the actual air velocities and distribution of temperatures expected to be found within the space.

Increasingly, CFD software is able to interact with other models, such as:

Boundary models.
Climatic models.
Building services models.
Energy consumption and CO2 generation models.
Radiant models.
Daylighting and lighting models.
CAD and BIM software.

[edit] Limitations

CFD can be a very useful tool in the right hands, and the output graphics are very persuasive and seductive. However results are highly dependent on the knowledge of the person setting up the model and interpreting the results. This is an increasing concern as CFD software becomes more straight-forward to use and so is more easily operated by people with little understanding of the mathematical model that underpins it.

If the input information is wrong, the output information will be as well. CFD is no substitute for common sense.

An important consideration in developing a CFD model is the generation of the grid of cells. The greater the number of cells, the more accurate the simulation will be, but the longer the model will take to run. In some parts of a space, using a large cell size may not have a significant impact on the results, however in sensitive areas, for example around complex boundaries or where there might be a large temperature difference between a boundary surface and the air next to it, it is important that cells are as small as is computationally practical. For example a very small cell size (a fine grid mesh) is necessary to properly simulate the downdraft that can be experienced next to a cold window. If such a downdraft is not simulated, the heat exchange between the window and the space it encloses will be underestimated.

In spaces where the surfaces enclosing the space are non-cartesian (ie they are curved, or at an angle rather than purely horizontal or vertical) it is important that the grid is body-fitted (ie that it follows the contours of the surfaces) rather than cartesian (in a series of steps), otherwise air velocity at the surface will be underestimated and so heat transfer between the space and its enclosing surfaces will be underestimated. This is particularly important where there is a large temperature difference between the surface and the air adjacent to it.

Where CFD is being used to predict user comfort within a space, it is important that both air temperature and radiant temperature are considered. CFD in itself only models air temperature and air velocity, however, around half of the contribution to our thermal comfort within buildings is dependent on radiant heat transfer, ie the temperature of the surfaces around us. Some CFD software is able to include radiant influences on the temperatures that will be felt by occupants.

[edit] Conclusion

In the right hands, CFD can make a useful contribution to understanding the likely performance of a building. However, using CFD properly is time consuming and can be expensive, and in many circumstances there may be other, more straight-forward analytical techniques that will be more appropriate.

In cases where there are unique or complex circumstances, and an understanding of fluid behaviour is critical to the success of the building, CFD can be the only option available.