Air flows in an open plan air conditioned office

Application Challenge 4-03 © copyright ERCOFTAC 2004

Comparison of Test data and CFD

The DOAP for this case are the temperature within the occupied zone (the lowest 1.8m of the space). The measured profiles of temperature are now compared with the results computed with the CFD1 and CFD2. Comparisons are made only for Case 2, an ACR of 2.0 per hour.

Table 5 shows values computed for the temperature at the 3 measurement locations from CFD1 design stage and CFD2 post occupancy stage. These values should be compared with the measurements which are listed in Table 2.

CFD1 Design Stage
Height (m)	A'	A	B
0.1	23.6	23.7	23.7
1.1	24.3	24.3	24.3
1.8	25	24.6	24.6
CFD2 Post occupancy
Height (m)	A'	A	B
0.1	23.9	23.7	23.4
1.1	25	25	25
1.8	25.2	25.2	25.2

Table 5 Calculated values of temperature (°C) through profiles at each of the 3 measurement locations

Firstly, Figure 8 shows a summary comparison between the computed temperature profiles and the measurements. It shows the values at each of the three locations at each of the three heights; the average temperature from each height (averaged over the three measurement locations) is joined by lines. Hence Figure 8 gives a first look at the fidelity of the CFD compared to the measurements. The CFD is seen to agree well with the measurements. More can be learnt from closer inspection.

Firstly, note that the value of the temperature within the occupied zone is computed to be in the range 23-25°C for both simulations, which agrees well with the observations. The volume-averaged air temperature of the space is determined by the total heat budget imposed on the space: the heat fluxes associated with the boundary conditions. Hence the differences in the mean temperature indicate sensitivity to the boundary conditions and imposed space heating. Taking the average of the measured values (ie here averaging only over the three profiles within the occupied zone) as a proxy for the total volume average, yields 24 °C. This is a little cooler than the values of 24.2 °C from CFD1 and 24.6 °C from CFD2 (both calculated also from the CFD results at the three heights at the three measurement locations). These differences reflect the differences that can arise through uncertainties in boundary conditions and also in space heating loads. That both simulations run a little too hot indicates that the space loads have been overestimated slightly. The differences between the mean temperatures from CFD1 and CFD2 indicate the sensitivity to boundary temperatures (resolution is unlikely to affect mean temperature). The boundary temperatures between the two cases are different by a comparable amount to the differences in mean temperature, namely 0.5 °C.

Secondly both the measurements and the predictions suggest that there is a slight stable stratification in temperature, despite the mechanically induced mixing by the swirl diffusers and the convection. The difference in temperature between 0.1m and 1.8m from both simulations compares well with the measured difference of about 2 °C, although the design stage simulations give a smaller temperature change of more like 1 °C. This might indicate that the higher resolution of the CFD2 simulations better represents the mixing processes and hence better reproduces the vertical temperature gradient.

Figure 8. Comparison between the measured and calculated temperatures. Black diamonds: measurements; red squares: CFD1 design stage; blue triangles: CFD2 post-occupancy. Smaller symbols values at three locations A, A’ and B; larger symbols joined by lines: average of values at three locations

Figure 9 shows contours of temperature from the design stage and post occupancy simulations, which indicates the spatial variability and patterns in temperature produced in the simulations. Figure 9 shows that in general there is more variation in air temperature near the ground (see the figure at z = 0.1m) because the cold air emitted by individual swirl diffusers has not mixed well with the environment. At higher temperatures the temperature is rather well mixed. Clearly the location of measurements at very low levels needs to be chosen with care. These general conclusions are also seen in the differences in the vertical temperature profiles at the three measurement locations (Table 5 for the CFD results and Table 3 for the measurements). The maximum spatial variation at the lowest level of 0.1m in both CFD simulations is something like 1.5 °C, which is larger than the variation of 1 °C seen in the measurements. It does not seem appropriate to read great significance into these differences. In the simulations the maximum differences are between the core of the cold swirl jets and the surrounding ambient. So any local disturbance (caused by furniture or geometry not resolved in the CFD) to the real flow will change this spatial variation.

In summary the simulations compare well with the measurements of the parameters that are critical from a practical point of view, namely the temperature in the occupied space. The main uncertainties seem to be in specifying boundary conditions on the thermodynamics and on the space heating loads from occupants, computers etc although there is an indication that resolution plays some role in determining the stratification computed by the CFD.