U p w a r d – A i r f l o w V e n t i l a t i o n S y s t e m
769
the current research is increasing the pressure coefficient module at the outlet opening
(point S in figure 11), where the air is sucked out.
The Cp is obtained by empirical models:
(1)
=
−
0
0.5 . . Ū
2
In Eq. (1), above,
0
(Pa) is the static pressure in the undisturbed flow and U (m/s) is the
free stream velocity calculated at building height and is the density. The negative
pressure coefficient (
p
), obtained at the outlet opening wall surface (Point S), of each of
the sawtooth roof models were analysed by using the Software Ansys CFX.
In order to evaluate the sawtooth roof geometry effect in ventilation efficiency (ventilation
ratio and internal flow pattern), firstly the pressure coefficient was measured in the
leeward outlet opening wall of each model. More details of these results are given in
section 5.
4. CFD Simulation and boundary conditions
The computational settings and parameters for this study are outlined and finally the
results are presented. At the end there is a discussion about the architectural variables
and the results.
4.1. Computational domain size and grid resolution
The computational domain and grid setting were selected according with the best
practice guideline by (Franke et al. 2007) and (Tominaga et al. 2008). The dimension of
the domain were W x D x H = 96 x 36 x 36m. To verify the accuracy of different mesh
sizes and observe the main internal flow feature, three types of meshes were used.
These settings are described in table 2
and the unstructured mesh is shown in figure 15.
Table 2. Mesh types for the leeward sawtooth roofs
Mesh name
Grid20
Grid40
Grid60
Mesh length scale on the model surface (m)
0.30
0.15
0.10
Mesh resolution on model surface
20
40
60
1
st
cell height above surfaces (m)
0.50
0.50
0.50
Vertical expansion factor
1.10
1.10
1.05
Mesh length scale in indoor space (m)
0.30
0.15
0.10
Total number of cells
608,813
1,063,218 8,261,294
The convergence criteria used on the coarse (Grid20), medium (Grid40) and fine grids
(Grid60) were the root-mean-square of normalised residual for all variables to be less
than
1 x 10
−4
(RMS). The CFD simulations of the sawtooth roof Model A1 on the three
grids were performed for the normal (0°) wind direction and velocity of 1m/s using the
RNG
k
-ε turbulence model. The pressure coefficients at a vertical line 1.40m away
from the centre of outlet opening surface (Fig. 13) were selected to compare the results
of the three grids. On the fine grid (Grid60), the value had 4% difference from the
other two grids at the leeward opening wall. Among the three grids, the maximum value
of the difference was 0.04 on the leeward opening wall.
