CORE Metadata, citation and similar papers at [Link].

uk
Provided by UTHM Institutional Repository

Simulation of Airflow in Lecture Rooms

Johnson Lim Soon Chong 1 , Adnan Husain2 & Tee Boon Tuan3
1
Faculty of Technical Education, 2 Faculty of Mechanical Engineering & Manufacturing
Kolej Universiti Teknologi Tun Hussein Onn, Batu Pahat, Johore, Malaysia.
3
Faculty of Mechanical Engineering,
Kolej Universiti Teknikal Kebangsaan Malaysia, Ayer Keroh, Malacca, Malaysia.

ABSTRACT

The purpose of this study is to perform simulation of airflow in lecture rooms using Computational Fluid
Dynamics (CFD) technique. Two lecture rooms with different location of air-conditioning units are chosen as
study fields for measurements and simulation purposes. The air distribution parameters addressed in this study
are air velocity and air temperature. Measurements were also done on the boundary parameters of the room such
as air-conditioning inlet and outlet conditions. The simulations were performed by using CFX software without
involving occupants. Two types of turbulence models were compared for suitability in lecture room modeling.
Validation of results through linear regression technique was performed using SPSS software. Based on the
stream plots, the zones below the air-conditioning units were identified to have inadequate airflow. In overall, the
RNG κ – ε Turbulence Model yields better prediction than Reynolds Stress Turbulence Model. Simulation can be
a useful technique to predict air distribution performance in a lecture room for a better thermal environment.

Keywords: Airflow, Simulation, Modeling

(1) INTRODUCTION Based on literature studies, there is less study of

room airflow modeling which involves lecture rooms.
Most of the students spend majority of their time Furthermore, majority of the CFD room modeling
in lecture rooms during lecture or tutorial. Air flow involves modeling of rooms such as cold storage room,
distribution from the air-conditioning system becomes telecommunication room and operating room where
important in this case, as it will determine whether only little emphasis is given to total air distribution
students perceive the air velocity and air temperature performance and occupant’s thermal sensation. Other
in the acceptable comfort range. studies involved rooms where important room
This study emerged based on a need to further parameter can be controlled [2-3]. However, lecture
investigate the current airflow distribution system in rooms (where KUiTTHO is concerned) are mostly
lecture rooms of KUiTTHO. The airflow in a typical occupied by students and their perceived thermal
lecture room in KUiTTHO is not uniform for the sensation are dependable on the thermal environment
current arrangement of air-conditioning units in the created by the air-conditioning system in the room.
room. It was also discovered that the position of air- Therefore, the study of the airflow modeling is crucial
conditioning units in the lecture rooms will affect the towards achieving desirable thermal environment in
air flow distribution in the room, which then causes the lecture room.
changes in thermal sensation among the occupants [1]. The main interest of this research study is to
There are some locations in the lecture room simulate airflow in the lecture rooms of KUiTTHO,
where occupants feel warm; and some other locations either with or without occupants. The main objectives
where occupants feel cold. This is mainly due to the are as follows:
improper selection of air-conditioning unit location (a) To identify and obtain important modeling
during the installation of the air conditioning system. parameters for valid CFD room modeling.
Therefore, a CFD study is necessary in order to (b) To evaluate air distribution performance in
investigate the airflow pattern and airflow distribution, lecture rooms through measurements and
which will help to identify the locations where air CFD techniques.
distribution is inadequate.
 (c) To compare and validate results obtained
from both the measurements and CFD
simulation techniques.

(2) METHODOLOGY

The scope of this study includes pilot and actual

measurements of the important air distribution
parameters and the simulation performed by using a
CFD software. The actual physical conditions in the
room are maintained without alterations during
measurements. The main procedures of this study are
as follows:
(a) Selection of suitable lecture rooms
(b) Performing actual measurement on important
air distribution parameters and boundary
parameters
(c) Performing CFD analysis on room air flow
through computer simulation
Before performing any physical measurements in
a typical lecture room, the lecture room is cleaned.
Eight rows of long chairs are arranged in both sides in
the room, according to a fixed arrangement, as defined
by the writer. The whole room is arranged accordingly.
A typical fixed arrangement for a lecture room is Fig.1 A Typical Fixed Arrangement for
shown in Fig.1. The exact location of each rows of Lecture Room.
long chairs, lecturer’s table, air-conditioning unit and
measurement points. All the other obstacles in the
lecture room including overhead projector (OHP),
OHP screen and extra long chairs are taken out of the
room during measurement. Curtains were closed in an
attempt to eliminate heat transfer through glass
window, as shown in Fig.2. All measurements were
done without the presence of occupants.
Two lecture rooms with different location of air-
conditioning units are chosen as study fields for
measurements and simulation purposes. In the case of
Lecture Room 1 (BK 17) and Lecture Room 2 (BK 23),
the air-conditioning units were installed at the side of
the rooms. Two rooms, each with different supply
location, will then be chosen for actual simulation
purpose. The dimensions for the mentioned lecture
rooms are standard with 9.64 m long, 7.11 m wide and
3 m height.

Fig.2 Closed Curtains to Minimize Heat

Transfer through Windows.
In order to assess the lecture room’s air
distribution performance using ADPI, the lecture room
has been divided into 20 zones, following the methods
used by Tee [1]. The zones are dotted lines that divide
the room into 20 zones, as shown in Fig.1. The points
of measurements are located at the center of each zone,
on a fixed coordinate of location.
Before performing measurement using
anemometer, a typical lecture room is let to stabilize
for one hour, with reference to the methodology used
by Tee [1]. The purpose is to make sure that the air-
conditioning system had stabilized in order to produce
stable airflow for measurements. After the room is
stable in one hour, the anemometer was let to stabilize
for 5 minutes in the room for environment adaptability
before taking the real data at each measurement point.
The time and height of measurements are
following the standard as described in ASHRAE
Standard 55-92 [4]. The probe of the anemometer was
set at the height of 1.1 meters for both pilot
measurement and actual measurement. As the point of
measurements is near long chair, the anemometer’s
probe is fixed at the long chair using adhesive tape.
For other areas, the anemometer is fixed at a normal
plastic chair. For a typical point of measurements, the
time to record readings is 3 minutes. The data taken Fig.3 Lecture Room with Side Location of Air-
for a particular point of measurements are air velocity Conditioning units in CFX.
and air temperature.

Simulation Procedures

The first step in CFD modeling is the model

building. The whole lecture room was drawn
geometrically following real physical measurement
data, as described in Appendix A. In addition, the
lecturer’s table, long chairs and two types of air -
conditioning units were also modeled according to
their own actual physical measurements in CFX. Fig.
3 shows a lecture room with side location of air-
conditioning units modeled in CFX. Lecture room
with back location of air-conditioning units as
modeled in CFX is also shown in Fig. 4. All the
simulations were also done without the presence of
occupants.
The second step in modeling is to define the
fluid domains. A fluid domain is the area where fluid
flow is defined. In CFX, The fluid used was set to “Air
at STP”. The reference pressure was 101325 Pa, the Fig.4 Lecture Room with Back Location of Air-
atmospheric pressure. The simulation type was steady Conditioning units in CFX.
state with stationary domain motion. For the fluid
models, either the RNG κ - ε Turbulent Model or Data Analysis Method
Reynolds Stress Turbulent Model were chosen. For
each room, these two models were used to identify a Upon collecting all the actual measurement
better turbulent model for room air flow prediction. data, the data was analyzed using a few methods. As
each room was divided into 20 zones with 20 points of
measurement, the Effective Draft Temperature (EDT)
was evaluated for each point using the air velocity and
air temperature obtained. ADPI was then evaluated for
each room. Upon choosing the two lecture rooms to be
modeled in CFX, the two rooms were modeled in CFX During the course of this study, there are a few
to obtained predicted values of air velocity and air assumptions being made. These assumptions are:
temperature. Based on these data, predicted EDT was (a) The air-conditioning systems studied are all in
obtained and predicted ADPI was then evaluated for good and stable operations.
each room. (b) The measurements and simulation are done in a
The actual results were compared with the completely confined room space, with all the
simulated results by using overall average error doors and windows closed.
calculations and absolute error. The equations are: (c) There is no construction leakage or noticeable gap,
which leads to outdoor air infiltration.
(d) The effects of physical condition of the room
Xi − Xi (floor construction, difference in location of glass
CFD exp ……(1)
E ABS = × 100 % windows and doors, etc.) are neglected in this
X i
exp study.
(e) The anemometer used is well calibrated and in
n i i good conditions during measurement.
∑ X CFD − X exp
(f) The airflow produced in the room studied is
E = i =1 × 100% … … (2) steady during measurements, where steady state
n i
∑ X exp conditions are assumed for simulation.
i =1 (g) The heat transfer through glass windows and
doors are neglected. Only the heat transfer
Where X can be either air velocity or air through wall and ceiling are considered.
temperature. | Xi CFD - Xi exp | is the absolute difference (h) Internal heat sources in the lecture room such as
between simulated values and actual measurement fluorescent lights are considered to have small
values for variable X. Linear regression analysis was effects and thus neglected in study.
also carried out to validate the results. Both the (i) The outside temperature is constant at 30 °C
Microsoft Excel 97 Spreadsheet Program and the during measurements.
SPSS Statistical Package, Release 10.0 were used for (j) Air velocity of 0.10 m/s is considered as still air
linear regression analysis. From the analysis, the and air velocity difference of below than 0.10 m/s
governing equation that links between the actual and as insensible [5] and small.
the simulated values can be determined. (k) ADPI of 60 to 69 is considered as unsatisfactory,
The residual squared value, R2 and the Pearson’s 70 to 79 as satisfactory and 80 and above as good
Correlation Coefficient, R can be determined and serve air distribution.
as a factor between the degree of linear relationship
between the two variables studied. The R values (3) RESULTS AND DISCUSSION
ranges from –1 to 1, where –1 indicates strong
negative linear relationship and 1 indicates strong A stream plot displays streamlines, which can be
positive linear relationship. If R is 0.8, it means that thought of as the paths of imaginary mass-less
80% of the data were explained by linear relationship. particles through the geometry in a steady-state
In order to interpret the various values of R, Guilford’s calculation. This is usually useful for showing the
suggested interpretations of R-values could be used, as particle paths through the geometry. In this study,
shown in Table 1. since the RNG κ – ε Turbulence Model is a better
model for lecture room model representation, the
Table 1 Guilford’s Suggested Interpretations for stream plots for lecture room BK 17 and BK 23 in
Values of R [6]. isometric view are obtained and shown in Fig. 5 and
R values Interpretation Fig. 6.
From the stream plots for lecture room BK 17 in
Less than 0.20 Slight; almost negligible relationship
Fig. 5, it can be observed that the cold air drops in the
0.20 to 0.40 Low correlation; definite but small
relationship front location of the room for air-conditioner no. 1 and
0.40 to 0.70 Moderate correlation; substantial air-conditioner no. 2. Upon dropping of air, mixing of
relationship air occurs in the front location of the lecture room as a
0.70 to 0.90 High correlation; marked relationship result of cold air recirculation. The middle of the room
0.90 to 1.00 Very high correlation; very dependable experienced good mixing of air. The back location of
relationship the room, particularly the area below the air
conditioning units, experienced poor mixing of air
where air flow is inadequate. This resulted in a 0.10 m/s. It was found that RNG κ – ε Turbulence
stagnant area at the back of the room, where relatively Model produced 85% of air velocity results with
higher temperature and lower air velocity values at difference of below 0.10 m/s, in both simulations in
location 17, 18, 19 and 20. lecture room BK 17 and BK 23. The Reynolds Stress
Turbulence Model had 70% and 80% of the results
with difference below 0.10 m/s in simulations of BK
17 and BK 23. In this case, the RNG κ – ε Turbulence
Model performs slightly well than Reynolds Stress
Turbulence Model.

Table 2 Percentage of Results with Absolute

Difference of Air Velocity Value of Below 0.10 m/s.
Lecture Turbulence Percentage of Results < 0.10
Room Model m/s (%)
BK 17 RNG κ –ε 85
Reynolds Stress 70
BK 23 RNG κ –ε 85
Reynolds Stress 80

Table 3 Percentage of Results with Absolute

Fig. 5 Stream Plots for RNG κ – ε Turbulence Difference of Air Temperature Value of Below 1 °C.
Model Simulation in Lecture Room BK17.. Lecture Turbulence Model Percentage of Results < 1 °C
Room (%)
BK 17 RNG κ –ε 100
Reynolds Stress 45
BK 23 RNG κ –ε 90
Reynolds Stress 65

For air temperature difference, it is shown in

Table 3 that for simulation in BK 17 using RNG κ – ε
Turbulence Model, all the difference of simulation
results with the actual values are below 1 °C, which is
small in value. Fo r simulation in BK 23, 90% of the
difference in simulation results are below 1 °C.
Comparing with the simulations using Reynolds Stress
Turbulence Model, there are only 45% and 65% of the
simulation results with difference below 1 °C.
Therefore, it can be concluded that RNG κ – ε
Turbulence Model predicts air temperature more
Fig. 6 Stream Plots for RNG κ – ε Turbulence accurately than Reynolds Stress Turbulence Model.
Model Simulation in Lecture Room BK 23. The comparison between overall average error
percentage in this study with the previous study
The validity of results can be judged in terms of conducted by Hoang [Link] [3] are shown in Table 4
the absolute difference, overall average error and Table 5. Only air velocity results were compared
percentage and the results from linear regression as it was the only relevant results with study conducted
analysis. In terms of difference of in air velocity, it by Hoang [Link] [3]. The comparison revealed that the
was assumed that the difference in air velocity of overall average error percentage for this study is
below 0.10 m/s as insensible and small. This higher than overall average error percentage by Hoang
assumption was made following the study done by [Link] [3].
Kubo [Link] [5]. This study also assumed that the
temperature difference of below 1°C is small.
Table 2 shows in summary the percentage of
results with absolute difference of air velocity is below
Table 4 Comparison of Overall Average Error 0.9 for each simulation results produced with RNG κ –
Percentage for Air Velocity with the Previous Study. ε Turbulence Model. It is therefore concluded that
Present Study Hoang high correlation and marked relationship exist between
[Link]
(1999)
simulation results and measurement results.
RNG RNG Reynol Reynol RNG κ – ε, This showed that the boundary conditions in the
Simulati κ – ε, κ – ε, ds ds in Cold model are defined correctly. The important boundary
on in in BK Stress, Stress, Storage conditions that should be defined for correct modeling
BK 23 in BK in BK Room
17 23
are the inlet-outlet conditions of the air-conditioning
17
Overall units and the heat transfer properties by the ceiling and
Average wall of the room model.
Error 34.87 32.81 72.62 53.31 28.5 The stream plots produced by CFD simulation
Percenta can serve as a useful tool to observe airflows in room.
ge (%)
It can also show regions with inadequate airflow that
leads to poor mixing of room air. Based on the stream
Table 5 Comparison of Overall Average Error plots, the area under the air-conditioning units showed
Percentage for Air Temperature in Present Study. inadequate air flow, which causes lower air velocity
Present Study and higher air temperature. In general, simulation can
Simulation RNG RNG Reynolds Reynolds be a useful technique to predict air distribution
κ – ε, κ – ε, in Stress, in Stress, in performance in a lecture room for a better thermal
in BK BK 23 BK 17 BK 23 environment.
17
Overall
Average 1.49 2.58 4.74 3.92 (5) REFERENCES
Error
Percentage [1] B.T. Tee, Effects of Airflow Towards Thermal
(%) Comfort in Lecture Rooms, [Link]. Thesis, Universiti
Teknologi Malaysia, 2001.
With regards to air distribution performance of [2] P. Depecker, et. al., “Study And Comparison Of
lecture rooms, both lecture room BK 17 and lecture Two Types Of Air Flow In Operating Rooms Using A
room BK 23 has the same measured ADPI value of 70, CFD Code”, Proceedings of the 7 th Int. Conference On
which is satisfactory in terms of air distribution Indoor Air Quality and Climate: Indoor Air 96, July
performance. The RNG κ – ε simulation predicted 1996, Nagoya Japan, Vol. 2, pp. 719 –723,1996.
lower ADPI value in lecture room BK 17 (ADPI of [3] M.L. Hoang, et. al., “Analysis of the Air Flow in
65), but higher ADPI value in lecture room BK 23 A Cold Store By Means Of Computational Fluid
(ADPI of 75). Dynamics”, International Journal of Refrigeration,
Vol. 23, pp. 127 –140, 1999.
(4) CONCLUSIONS [4] ASHRAE.” ASHRAE Standard Thermal
Environment Conditions for Human Occupancy.”
The simulation of airflow in lecture rooms was Atlanta: (55-1992R). 2001
successfully carried out. Measurement results were [5] H. Kubo, [Link]., “Human Response To The
used to compare with the simulation values. This study Different Body Surface Area Exposed To Airflow”,
showed that the RNG κ – ε Turbulence Model predicts Proceedings of the 7th Int. Conference On Indoor Air
better than Reynolds Stress Turbulence Model and Quality and Climate: Indoor Air 96, July 1996,
therefore is suitable as turbulence model-to-model Nagoya Japan, Vol. 2, pp. 553 – 558, 1996.
lecture room. Majorities of over 80% of the simulation [6] R.C. Sprithall, Basic Statistical Analysis(4th
values are found to have small difference with actual Edition), Massachusetts, Allyn and Bacon, 1993.
measurement values in terms of air velocity and air [7] T.H. Kuehn, [Link]., Thermal Environment
temperature. Overall average error percentage of Engineering, New Jersey, Prentice-Hall, 1998.
simulations done in this study is all over 30%, which is [8] M. Loomans, The Measurement and Simulation
higher than the overall average error percentage of Indoor Air Flow, Ph.D. Thesis, Eindhoven
produced by previous study. University of Technology,1998.
Nevertheless, linear regression analysis showed [9] M. Ramey, Application of Computational Fluid
good agreement between simulation values and Dynamics to Indoor Room Airflow, [Link]. Thesis,
measurement values. This is evident from the values of Oklahoma Christian University of Science and Arts,
Pearson’s correlation coefficient, R of between 0.7 and 1994.