External Flow Comparison Using CFD
China Project  Beijing 3/7/2000
Sephir Hamilton, Jeff Huang, Nobukazu Kobayashi, Dong Luo
Abstract
This document evaluates whether the study of natural ventilation
in a building can be separated into two parts: indoor airflow and
outdoor airflow. This decoupling allows CFD users to use necessarily
coarse grids for exterior modeling and finer grids for interior
modeling The two models are then coupled by pressure boundary conditions
at the windows. Modeling a complete external and internal flow system
with a fine enough grid for internal flow creates files that are
too computationally intensive for available computers.
The results of the study compare pressure distributions
around objects modeled as pure blockages (no openings) and as partial
blockages with various windowtowall ratios. Errors in pressure
distribution up to 20% are realized between the pure blockage and
the highest windowtowall ratio case. Using decoupled models is
not acceptable for precise quantitative flow simulation. For conceptual
design, however, the separation of indoor airflow and outdoor airflow
for natural ventilation design is acceptable.
Models
Two separate experiments each show similar results. One
experiment uses a simple rectangular blockage, the other simulates
the facade of an actual building design (China Project Design 
Beijing).
The first experiment (Hamilton and Huang) models
a pure rectangular blockage in 2 m/s airflow. Then it models a blockage
with an open interior, walls, and exterior windows (3 cases each
with different windowtowall ratios  figure 1).


Figure
1 The blockage allows air through the windows (perpendicular
to the wind). The interior partitions simulate an actual
apartment's flow restriction. 15% (case 1), 11.1% (case
2), and 6.25% (case 3) windowtowall areas were each modeled.
The results from each model are compared to the 0% window
area model.

Figure
1a  View of the model with the entire modelspace



The second experiment (Kobayashi and Luo) models a pure blockage
in the shape of the Beijing building design (China Project, 2000)
in 2 m/s airflow. Then it models the same blockage with 30% windowtowall
ratio (no internal obstructions  figure 2) and with 30% windowtowall
ratio (with internal obstructions/walls  figure 3).
Figure
2  The blockage with 30% windowtowall area, no obstructions





Figure
3  The blockage with 30% windowtowall area, with internal
obstructions (walls  in white)

Results
Experiment 1 (Hamilton and Huang)
The base case without the windows yielded a pressure difference
between the windward and leeward sides of 4.259Pa (windward and
leeward pressures equaled 5.274Pa and 1.015Pa respectively) These
values were taken at the centerline. The pressure difference in
case 1 (15% windowtowall area) revealed a slightly lower pressure
difference of 3.809Pa at the centerline, and 3.012Pa locally at
the center of one pair of windward/leeward windows. The case 2 (11.1%
window) pressure differences were 3.896Pa and 3.041Pa at the centerline
and the window centerline respectively. Case 3 (6.25% window) had
pressure differences of 3.863Pa at the centerline and 3.009Pa at
the window centerline. Graphic and tabular results are shown below.
Table
1  Results of pressure study for Experiment 1

Windward Pressure [Pa]

Leeward Pressure [Pa]

Pressure Difference [Pa]

% Difference of Pressure Difference

Base Case

5.274

1.015

4.259

N/A

Case 1 (centerline)

5.074

1.265

3.809

10.6

Case 1 (window centerline)

4.607

1.595

3.012

29.3

Case 2 (centerline)

5.069

1.173

3.896

8.52

Case 2 (window centerline)

4.577

1.536

3.041

28.6

Case 3 (centerline)

5.035

1.172

3.863

9.3

Case 3 (window centerline)

4.701

1.529

3.009

25.5

Figure
4  Pressure distribution around the pure blockage (base
case)





Figure
5  Resulting velocity vectors for case 1

Figure
6  Pressure distribution for case 1





Figure
7  Pressure distribution for case 2

Figure
8  Pressure distribution for case 3



Experiment
2 (Kobayashi and Luo)
In Case1 (without windows), the pressure difference between the
front and the back of the building is approximately 2040% higher
than that of Case2 (with windows and without inner walls). However,
in Case3 (with windows and inner walls), the pressure difference
between the front and the back is almost same as that of Case1
(without windows). The inner walls work as resistance to air flow.
The results imply that windows do not make so much difference to
the calculation of the pressure around buildings. The results are
shown graphically and in tabular form below.

Windward Pressure

Leeward Pressure

Pressure Difference

% deviation from base case

Case 1 (base case)

8.3  7.8Pa

0.8 Pa

9.1  8.6

N/A

Case 2 (no internal walls)

6.5  5.0

0.5  1.2

7.7  5.5

~25%

Case 3 (with walls)

7.3  6.3

0.8  1.8

9.1  7.1

~8%

Figure
9  Velocity distribution for case 1 (base case)





Figure
10  Pressure distribution for case 1

Figure
11  Velocity distribution for case 2





Figure
12  Pressure distribution for case 2

Figure
12  Velocity distribution for case 3





Figure
13  Pressure distribution for case 3

Conclusion
Using decoupled models is not acceptable for precise
quantitative flow simulation. For conceptual design, however, the
separation of indoor airflow and outdoor airflow for natural ventilation
design is acceptable.
One can successfully decouple the exterior and
interior flow for a building exposed to wind to study natural ventilation
for conceptual design. The process, however, produces errors in
the magnitude of 1020% for pressure which is used to recouple
the exterior and interior flows.
back to top
