Best Practice Advice AC4-03
Air flows in an open plan air conditioned office
Application Challenge 4-03 © copyright ERCOFTAC 2004
Best Practice Advice for the AC
Key Fluid Physics
This Application Challenge consists of a flow through an internal office space. The air within the space is stably stratified. The flow is driven by
• Mechanically forced flow from the floor-mounted swirl diffusers, which lead to momentum jets that mix with the surrounding stably stratified air
• Buoyancy driven flow arising from the space heating due to internal loads, such as occupants and computers, which are distributed through the lowest 1.2m above the floor. This heated volume then mixes convectively with the stably stratified layer aloft.
Across most of the room space, including the occupied zone, the mean velocities and turbulence levels are low, although they are higher near the source of the swirl diffusers.
The competency of the CFD in this AC is judged by the ability to model the temperature and air speeds within the occupied zone (defined to be the lowest 1.8m above the floor) given the value of the supply flow rate. These parameters are considered here. Other parameters that may be of interest are air quality (which is usually judged through age of air) radiant temperature and air moisture content; these quantities are not considered here. The DOAPs are controlled by the mean flow through, and the mixing within, the space, since these determine the eventual stratification in the air and hence temperature in the occupied zone.
The UFRs relevant to the flow are:
UFR 3-10 Plane wall jet : For this particular AC there are no wall jets, although there other applications of CFD to office spaces might well contain wall jets
UFR 4-09 Confined buoyant plume: the buoyantly forced component of the flow has dynamics in common with this UFR; although there is the obvious difference that in the AC considered here, the source of heat is distributed in the horizontal and so distinct plumes are not resolved in the RANS modelling
UFR 4-11 Simple room flow: the component of flow in the present AC that is forced by the swirl diffusers has similar dynamics to UFR 4-11. A difference is that, in the 4-11 case, the mechanically forced flow reaches the ceiling and spreads as a wall jet; here the jets do not penetrate far above floor level. (For this reason UFR 3-10 Plan Wall Jet is not relevant here).
An important aspect of the flow not captured in any of the UFRs is the interaction between the buoyant and mechanical components of the motion. The swirl jets inject cool air near the ground, which mixes with the air that is being warmed by the internal loads. The mixing of these two air masses is thought to control the eventual vertical temperature distribution in the room and hence the competency of the CFD in this AC.
The following uncertainties make high fidelity CFD difficult:
• Poor specification of the internal heating loads caused by human occupancy, computers, etc, within the space. The location, representation (either as a distributed or discrete point sources) and magnitude of these sources requires specification. There are industry standards for specifying the magnitude of these loads. The advice is to follow these standards. Since these loads are a large fraction of the total load, they are likely to influence comparisons between the CFD and measurements as was the case in this AC.
• Furniture, occupants and fine details of geometry within the space are not typically resolved. Whilst these might change the flow field locally, it is not expected to change the temperatures within the occupancy zone. The DOAPs are required to be satisfied irrespective of furniture position as was the case in this AC.
• Details of the inlet flow emanating from the swirl diffusers are not typically resolved explicitly. Instead, following BPA under UFR 4-11, the swirl diffusers are modelled, here as a vertical jet surrounded by 4 outer, angled, jets. The fraction of flow through the central jet, and the angles of the outer jets are varied to match the flow and temperature patterns from a single diffuser supplied by the manufacturer. (See UFR 4-11 documentation) The objective of this modelling is to represent the effect of the jet on the whole space, rather than represent the fine detail of the near field.
Computational Domain and Boundary Conditions
Lessons learnt from the present Application Challenge:
• The geometry of the domain is inherently 3d and not symmetric, and so has to be modelled completely. Steady state approximation is assumed. (In some applications, such as fire growth in the office space, unsteady effects are important.)
• It is important to represent carefully the flow characteristics at the swirl diffusers so that the mixing of the incoming air is well represented.
• Fluid-dynamical boundary conditions are the standard no slip conditions applied in this Application Challenge using standard wall functions.
• Thermal boundary conditions are also required on all solid surfaces either as specified temperatures (with or without a resistance representing conductive heat transfer) or specified heat fluxes. These are obtained either from observations or additional modelling. The results for the DOAPs are likely to be sensitive to these boundary conditions.
Discretisation and Grid Resolution
• The AC and UFR 4-06 used the SIMPLE solver. First or second order differencing schemes, such as upwind-types or MARS, are recommended, and second order differencing is to be preferred.
• The first grid point at a y+ within appropriate limits for use of wall functions
• UFR 4-11 indicates that good resolution of the inlet and outlet jets is required (requiring between 50,000-100,000 cells across the whole domain in the case considered there). Grid resolution for case studies depends strongly on the particular case.
• Accurate thermal boundary conditions are required
• The standard k-epsilon turbulence model was adequate for UFR 4-11 and UFR 4-14, with the RNG k-epsilon model working a little better for the buoyant plumes in UFR 4-14
• A buoyancy correction to the k-epsilon model is required to handle correctly the buoyant plumes. Options on how to do this are described in UFR 4-09
• The use of wall functions is generally acceptable since there are no strong wall jets in this AC. (When wall jets are present, a two layer approach is necessary, see UFR 3-10)
• The thermal boundary conditions on the walls need to be represented carefully as they can contribute to the heat balance of the space and hence change the temperature in the occupied zone
• In some office space applications, solar radiation is a primary driver for surface and air temperature distributions, and requires additional representation.
Recommendations for Future Work
• There is a pressing need for more details measurements against which to compare the simulations in this AC. When compared to the measurements in the UFRs discussed here, the amount of data available for this AC is limited. It is desirable to measure the mean flow and temperature across the occupied zone in a fully active office space. There are difficult practicalities to be overcome in achieving this aim. In addition the thermal boundary conditions need to be measured carefully.
• Both UFR 4-11 and UFR 4-14 suggest the use of higher order turbulence models to compute the flows. So it would also be appropriate to repeat this AC with higher order closure.
© copyright ERCOFTAC 2004
Contributors: Isabelle Lavedrine; Darren Woolf; Stephen Belcher - Arup
Site Design and Implementation: Atkins and UniS