See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237459502

Modeling of Ceiling Fan Based on Velocity Measurement for CFD Simulation of

Airflow in Large Room

Article  in  Journal of Environmental Engineering (Transactions of AIJ) · January 2004

DOI: 10.3130/aije.70.41_3

CITATIONS READS

9 825

4 authors:

Yoshihisa Momoi Kazunobu Sagara

University of Fukui Osaka University
30 PUBLICATIONS   26 CITATIONS    60 PUBLICATIONS   792 CITATIONS   

SEE PROFILE SEE PROFILE

Takamitsu Yamanaka Hisashi Kotani

Carnegie Institution for Science Osaka University
321 PUBLICATIONS   4,702 CITATIONS    82 PUBLICATIONS   351 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IEA-EBC Annex 62 Ventilative Cooling View project

All content following this page was uploaded by Takamitsu Yamanaka on 15 August 2014.

The user has requested enhancement of the downloaded file.

 Modeling of Ceiling Fan Based on Velocity Measurement for CFD
Simulation of Airflow in Large Room

Y. Momoi1, K. Sagara1, T. Yamanaka1 and H. Kotani1

1
Osaka University, Graduate School of Eng., Dept. of Architectural Eng., 2-1 Yamadaoka, Suita, Osaka, Japan
Email: momoi@arch.eng.osaka-u.ac.jp http://www.arch.eng.osaka-u.ac.jp/~labo4/

Summary: This study examines the way of utilizing a ceiling fan for airflow control in a large air-conditioned
room. Although it seems that CFD simulation is useful in predicting the airflow around a ceiling fan, modeling
of a ceiling fan as a body of rotation is very complicated. Therefore, in this study, airflow of a ceiling fan is
modeled as boundary conditions of air velocity data measured near the ceiling fan. In this paper, the measured
airflow pattern around a ceiling fan is compared with the CFD simulation result using the airflow model of the
ceiling fan, in order to examine the validity of the airflow model. The CFD result was in good agreement with
the measurement result concerning the average of air velocity.

Keywords: Ceiling fan, CFD, Boundary condition, Measurement data, Large room
Category: Modeling techniques

1. Introduction φ/3.

This study examines the way of utilizing a ceiling

52
47
fan [1, 2, 3] in a large air-conditioned room for the
dissolution of an indoor vertical temperature φ/56
gradient in winter and the prevention of draft in

/7/
summer. Although it seems that Computational
Fluid Dynamics (CFD) simulation is useful in

72
predicting the airflow around a ceiling fan,
modeling of a ceiling fan as a body of rotation is φ02.
φ/1..
very complicated. Since the number of divisions of
the space increases, calculation load becomes heavy. Cjct_rgml_lbApmqqQcargmlTgcu
As for complex-shaped supply openings, some
techniques of modeling and decreasing calculation
load are proposed by Nielsen [4]. In this study,
these techniques will be applied to a ceiling fan,
/0.Ň
and airflow of a ceiling fan is modeled as the
boundary conditions based on velocity data
measured near the ceiling fan.
In this paper, the measured airflow pattern around Nj_lTgcu Slgr8Ykk[
the ceiling fan is compared with the CFD Fig.1 Ceiling fan
simulation result using the airflow model of the
ceiling fan, in order to examine the validity of the
airflow model. @jmu+mddQspd_ac
8U, V, W, k, ε

2. Airflow model of ceiling fan 0. QsargmlQspd_ac

A ceiling fan (Matsushita Electric Industrial, F- 0.

8U, V, W, k, ε
/2.
M131H-W) used in this research is shown in Fig.1. /2.
Slgr8Yak[
In the CFD simulation of an indoor airflow induced
&/'?gpdjmuBgpcargml8Snu_pb
by this ceiling fan, two virtual planes whose each
side length is 140 cm are assumed at the position of
/2.
20cm above and below the ceiling fan. Measured /2.
values of air velocity and turbulent parameters near 0.
QsargmlQspd_ac

the ceiling fan are given as boundary conditions for 8U, V, W, k, ε

0.
airflow of the ceiling fan (Fig.2). The influence of a @jmu+mddQspd_ac
ceiling fan on an indoor airflow can be taken into 8U, V, W, k, ε
consideration in the CFD simulation. It becomes Slgr8Yak[

unnecessary to simulate correctly the space between &0'?gpdjmuBgpcargml8Bmulu_pb

two planes where airflow is complicated in reality, Fig.2 Schematic diagram of airflow model of ceiling fan
and calculation load can be reduced.
3. Experimental set-up
The experiments were carried out in the large 1.. /3.. 1..
experimental room. The iron framework with
dimensions of 3.0m x 1.5m x 3.0m was constructed Acgjgle
and the ceiling fan with a rotational radius of 65 cm
was installed in the height of 2 m above the floor @c_k @c_k

63.
level (Fig.3). Measurement for 30 seconds was

/3..
X
performed by 10Hz of sampling frequencies using Acgjgle
the 3-dimensional ultrasonic anemometer (KAIJO, D_l W
V
WA-590), and the average value of each air

203.
velocity component was calculated. Measurement

/3..
for 20 seconds was carried out by 250Hz of

0...
sampling frequencies using hot-wire anemometer of 1BSjrp_qmlga
?lckmkcrcp
constant temperature type (
5µm tungsten,
KANOMAX JAPAN, 0251 R-T5), and turbulent Rp_tcpqcp
/3
..

kinetic energy k [m2/s2] and energy dissipation of it Kc_qspgleQcargml

 [m2/s3] were computed from the auto-correlation Djmmp 1...
of measured air velocities. Measurement was Slgr8Ykk[
Fig.3 Experimental equipment
performed at intervals of 5cm or 10cm. The
measuring field was set between 80 cm above the
ceiling fan and 120 cm below it, and horizontally
Kc_qspglenmglrqugrfsjrp_qmlga
up to 120 cm from the rotational center of the _lckmkcrcp8/54nmglrq
ceiling fan at each point of the cross section passing Kc_qspglenmglrqugrfsjrp_qmlga
XYak[
through the center of the ceiling fan (see Fig.4). _lbfmr+ugpc_lckmkcrcp864nmglrq
Kc_qsp_`jc?gpTcjmagrw
The speed of rotation was fixed at 160 rpm. As the Amknmlclr
6.
6.
ceiling fan can generate either upward or downward
flow depending on the rotational direction, the
measurements were conducted under the two
experimental conditions that airflow directions 2.
2.
from the ceiling fan were upward and downward
c p
cr

0.
0.
ck gpc

(Table 1).
mk
?l r+u
Fm

.
.
4. Boundary Condition 43
+0.
+0.
First, regression curves were obtained from the Kc_qsp_`jc?gpTcjmagrw
measured velocity data on the line (see Fig.5). Amknmlclr
+2.
+2.
Second, the boundary surface set at the position of
20cm above and below the ceiling fan was divided
into the mesh of 1cm interval, and the values of air
velocity components and turbulent parameters in +6.
+6.
cr ga
mk qml

the central point of each mesh were computed using

cp
ck r p _

the regression curves. Finally, the values given to a

?l Sj
1B

boundary surface were acquired by averaging them +/0.

VYak[
in 10cm mesh. The application section of a
regression formula was divided considering the . 0.
0. 2. 4. 6. /.. /0.

rotational radius (X=65cm) and the position where Fig.4 Measuring points
measured air velocity was the maximum value
(upward: X=35cm, downward: X=45cm). All Table 1 Experimental condition
regression formulae were taken as a cubic function.
In this way, the boundary conditions on the meshed
?gpdjmu
plane of CFD simulation are determined. The bgpcargml
boundary conditions under downward condition Bmulu_pb Snu_pb
and upward condition are shown in Fig.6.
Qnccbmd
pmr_rgml /4.pnk
 V+amknmlclrmdTcjmagrw

W+amknmlclrmdTcjmagrw

X+amknmlclrmdTcjmagrw
.,0 / .,3

.,6 .
.
/.0.1.2.3.4.5.6.7./..//.
X;)0.ak
/.0.1.2.3.4.5.6.7./..//.
+.,3

UYk-q[
TYk-q[
SYk-q[ +.,0 .,4
+/
+.,2 .,2
+/,3
+.,4 .,0
+0

+.,6 ./.0.1.2.3.4.5.6.7./..//.
+0,3
VYak[ VYak[ VYak[
?gpdjmubgpcargml8Bmulu_pb

.,0 / .,3
V+amknmlclrmdTcjmagrw

W+amknmlclrmdTcjmagrw

X+amknmlclrmdTcjmagrw
.,6 .
.
/.0.1.2.3.4.5.6.7./..//.
/.0.1.2.3.4.5.6.7./..//.
X;+0.ak

+.,3

UYk-q[
TYk-q[
SYk-q[

+.,0 .,4
+/
+.,2 .,2
+/,3
.,0
+.,4 +0
/.0.1.2.3.4.5.6.7./..//.
.
+.,6 +0,3
VYak[ VYak[ VYak[

 


 
V+amknmlclrmdTcjmagrw

W+amknmlclrmdTcjmagrw

.,4 .,0

X+amknmlclrmdTcjmagrw
/,0
.,2 /
.
/.0.1.2.3.4.5.6.7./..//.
X;)0.ak

.,6
.,0
+.,0

UYk-q[
TYk-q[
SYk-q[

.,4
.
/.0.1.2.3.4.5.6.7./..//. +.,2 .,2
+.,0
+.,4 .,0
+.,2
.
+.,4 +.,6 /.0.1.2.3.4.5.6.7./..//.
+.,0
+.,6 +/ +.,2
VYak[ VYak[ VYak[

?gpdjmubgpcargml8Snu_pb

.,4 /,0
V+amknmlclrmdTcjmagrw

W+amknmlclrmdTcjmagrw

X+amknmlclrmdTcjmagrw

.,0
.,2 /
.
X;+0.ak

.,0 /.0.1.2.3.4.5.6.7./..//. .,6

+.,0
UYk-q[
TYk-q[
SYk-q[

.,4
.
/.0.1.2.3.4.5.6.7./..//. +.,2 .,2
+.,0
.,0
+.,4
+.,2 .
+.,4 +.,6
/.0.1.2.3.4.5.6.7./..//.
+.,0
+.,6 +/ +.,2
VYak[ VYak[ VYak[
2



 
Fig.5 Regression curves
VYak[
5. +43+33+23+13+03+/3+3 3 /3 03 13 23 33 43 5.
. .,/3
IglcrgaClcpew

IglcrgaClcpew
iYk0-q0[

iYk0-q0[

3. .,..3 3. .,/.

1. .,./. 1. .,.3

.,./3 .
WYak[

WYak[

/. /.
+43+33+23+13+03+/3+3 3 /3 03 13 23 33 43
. . VYak[
rsp`sjclaciglcrgaclcpew

rsp`sjclaciglcrgaclcpew

VYak[
+/. +/.
Bgqqgn_rgmlp_rcmd

Bgqqgn_rgmlp_rcmd

+43+33+23+13+03+/3+3 3 /3 03 13 23 33 43
. .,1
εYk0-q1[

εYk0-q1[

+1. +1.
.,./ .,0
+3. +3.
.,.0 .,/

+5. .,.1 +5. .

+5. +3. +1. +/. . /. 1. 3. 5. +5. +3. +1. +/. . /. 1. 3. 5. +43+33+23+13+03+/3+3 3 /3 03 13 23 33 43
VYak[
VYak[ VYak[
X X
X;)0.ak X;)0.ak
V V
/k-q /k-q

VYak[
5. +43+33+23+13+03+/3+3 3 /3 03 13 23 33 43 5.
. .,./3
IglcrgaClcpew

IglcrgaClcpew
iYk0-q0[

iYk0-q0[

.,.3
3. 3. .,./.
.,/.
.,/3 .,..3
1. 1.
.,0.
.,03 .
WYak[

WYak[

/. /. +43+33+23+13+03+/3+3 3 /3 03 13 23 33 43
VYak[
rsp`sjclaciglcrgaclcpew

rsp`sjclaciglcrgaclcpew

. .
VYak[
Bgqqgn_rgmlp_rcmd

Bgqqgn_rgmlp_rcmd

+/. +/.
+43+33+23+13+03+/3+3 3 /3 03 13 23 33 43
. .,..1
εYk0-q1[

εYk0-q1[

+1. +1.
.,0 .,..0
+3. .,2 +3.
.,../
.,4
+5. .,6 +5. .
+5. +3. +1. +/. . /. 1. 3. 5. +5. +3. +1. +/. . /. 1. 3. 5. +43+33+23+13+03+/3+3 3 /3 03 13 23 33 43
VYak[
VYak[ VYak[
X;+0.ak X;+0.ak


 


  2



 
Fig.6 Boundary conditions
5. CFD simulation set-up CFD simulations were carried out with the
As shown in Fig.7, CFD simulation of the airflow commercial CFD code (STREAM ver.5, Software
in experimental room was performed using the Cradle co. Ltd. [5]). The standard k-
two-equation
above-mentioned boundary condition for the ceiling turbulence model and SIMPLEC algorithm were
fan based on measured value. The size of applied with standard log-law wall function and
simulation space was determined by preliminary finite-volume resolution. A computational grid was
simulation. It is large enough to avoid the influence set up with 10cm interval.
of a wall on the airflow near the ceiling fan.
6. Comparison between the
. measurement and CFD result
..
/3
_gpdjmukmbcjmdacgjgled_l
6.1 Air velocity distribution
The measurement and CFD result of air velocity
/0..

under downward condition and upward condition

203.

X are shown in Fig.9.

V
(1) Downward airflow condition
W In both measurement and CFD result, it is seen
/3...
that the airflow goes toward the rotational center of
SlgrăYkk[
the ceiling fan (X=0cm) at Z=-20cm (blow-off side).
Fig.7 Simulation space
Although the maximum downward air velocity is
about 2m/s near X=45cm, air velocity falls
suddenly outside it and becomes almost 0m/s
outside the rotational radius of the ceiling fan
Acgjgke (X=65cm). At Z=-40cm, -80cm, and -120cm in the
measurement result, it turns out that the direction of
/0.. @c_k @c_k
blow-off airflow is changing outward gradually as
203. it becomes far from a ceiling fan. There seems to be
63. 1../3.. 1.. ?gpdjmukmbcjmd
3.
acgjgled_l&qsargmlqspd_ac(' the influence of the floor at 200cm under the ceiling
2.. 1..
3. ?gpdjmukmbcjmd
fan. However, in the CFD result, such a change of
/6.. /2.. acgjgled_l&`jmu+mddqspd_ac(' airflow direction observed in the measurement
X
result can’t be seen. The downward flow was struck
V on the floor and changed the upward flow near the
W
Djmmp
rotational center in the measurement result. The
Slgr8Ykk[
(Slbcpbmulu_pb_gpdjmuamlbgrgml,Ufclrfc_gpdjmubgpcargmlgq
upward airflow caused by rebounding of the
snu_pbamltcpqcjw*rfcsnncpqspd_acgqĤ@jmu+mddQspd_acĥ downward flow was not observed at Z=-120cm
_lbrfcbmulcpqspd_acgqĤQsargmlQspd_acĥ,
Fig.8 Cross section around ceiling fan near the rotational center in the CFD result. The
insufficient grid division in the CFD field might
cause this inconsistency.
The cross section around the ceiling fan is shown (2) Upward airflow condition
in Fig.8. Beams were arranged with at intervals of In both measurement result and CFD result, in Z=
150cm. The airflow model of the ceiling fan was 20cm (blow-off side), the maximum upward air
installed in the middle between two beams. On velocity is about 1 m/s near X=45cm. Outward flow
account of the restriction of simulation software, is observed as a whole. Outside X=65cm, air
the boundary conditions of the blow-off and suction velocity was almost 0cm/s. At the height of Z=
airflow of the ceiling fan could not be given on one 40cm and 80cm, it turns out that the airflow blown
surface respectively, but the same boundary by the ceiling fan has spread outside since the peak
conditions are given to the upper and lower sides of of air velocity distribution is moving outside while
a box with a height of 5cm (the inside of the box upward flow becomes weaker. In a suction side, the
was excluded from the simulation space). That is, horizontal airflow changed into the perpendicular
for example under downward airflow condition, the airflow as the airflow approaches the rotational
blow-off surfaces were set at the position of 180cm center of the ceiling fan, and the airflow was sucked
(Z=-20cm) and 185cm (Z=-15cm) from the floor, toward the blades of the ceiling fan. In the outer
and the suction surfaces were set at 215cm side of the ceiling fan, the horizontal suction flow
(Z=+15cm) and 220cm (Z=+20cm) from the floor. that goes to the center of the ceiling fan can be seen.
Although the iron framework was constructed The influence of suction toward the ceiling fan
around the ceiling fan in the experiment, it was not becomes small as it becomes far from the ceiling
taken into consideration in the simulation since it fan. Concerning the suction flow of the ceiling fan,
appears that the framework does not affect the the CFD and measurement results are enough in
airflow near the ceiling fan. Simulation field was agreement and it would be said that it is analyzable
assumed to be isothermal so that temperature field with sufficient accuracy.
was not analyzed.
 Z [cm]
0.50
80 0.45
: Experiment (downward)
0.40 [Z=-40cm]
40 0.35

k [m2/s2]
20 : CFD (upward)
0.30
[Z=40cm]
0
0.25
-20 : Experiment (upward)
0.20
: CFD (downward) [Z=40cm]
-40 0.15 [Z=-40cm]
0.10
-80 0.05
0
0 10 20 30 40 50 60 70 80 90 100 110 120
-120
X [cm]
Fig.10 Turbulent kinetic energy distribution

1 m/s 1.4

1.2 : Experiment (downward)

X [cm]
[Z=-40cm]
0 20 40 60 80 100 120 0 20 40 60 80 100 120
1.0
Experiment CFD Simulation : CFD (upward)
ε [m2/s3] 0.8
[Z=40cm]
(1) Downward airflow condition : CFD (downward)
Z [cm] [Z=-40cm]
0.6 : Experiment (upward)
[Z=40cm]
80
0.4

40
0.2

20 0
0 10 20 30 40 50 60 70 80 90 100 110 120
0
X [cm]
-20 Fig.11 Dissipation rate of k distribution
-40
0.6
: Experiment (downward)
-80 [Z=-40cm]
0.5

-120 0.4
: CFD (downward)
[Z=-40cm]
l [m]

0.3
: Experiment (upward)
1 m/s
[Z=40cm]
0.2
X [cm]
: CFD (upward)
[Z=40cm]
0 20 40 60 80 100 120 0 20 40 60 80 100 120 0.1
Experiment CFD Simulation
(2) Upward airflow condition 0
Fig.9 Air velocity distribution 0 10 20 30 40 50 60 70 80 90 100 110 120
X [cm]
Fig.12 Turbulent length distribution
6.2 Distribution of turbulent kinetic
energy and its dissipation rate The CFD and measurement results of turbulent
Turbulent kinetic energy can be calculated as kinetic energy distribution and its dissipation rate
follows: distribution are shown in Fig.10 and 11,
k = 1 × 3 u'
2 2 ( 2
) respectively. The CFD result was compared with
the measurement result in blow-off side (at Z=-
and its dissipation rate is given by: 40cm under downward airflow condition and at
3
k2 Z=40cm under upward airflow condition). In the
= CD
l measurement result, they have tendency to become
Turbulent length is expressed as follows: large significantly at X=50cm to 55cm. On the
1
other hand, although the maximum values in the
l = CD4
CFD result have appeared in the almost same

is integral length and obtained from the auto- position as those in the measured result, they are
correlation of measured air velocities: smaller than the measured values. Since boundary
∞ ( ) ( )
= u ∫0 u' t u' t2+ d conditions are equalized among the width of a mesh
u'( t )
 (10cm), it is considered that non-uniform Acknowledgements
distribution like the measurement result is not The authors would like to thank Machiko Kuise
observed in the CFD result. In the other positions, for her help in carrying out the measurements and
the values in the CFD result are larger than those in CFD simulations. The provision of the ceiling fan
the measurement result. As a cause, the production from Matsushita Ecology System Co., Ltd. is
of turbulent kinetic energy due to air velocity gratefully acknowledged.
distribution may have been estimated too greatly.
Furthermore, since turbulent kinetic energy and Nomenclature
its dissipation rate distribution of the CFD k = turbulent kinetic energy [m2/s2]
simulation differ from those of the measurement in u' 2 = mean-square fluctuating component of
this way, it is thought that the difference would velocity [m2/s2]
induce air velocity distribution near the rotational = dissipation rate of turbulent kinetic
center of the ceiling fan. energy [m2/s3]
6.3 Turbulent length distribution CD = constant (0.09)
The CFD and measurement results of turbulent l = turbulent length [m]
length distribution are shown in fig.12. Turbulent = integral length [m]
length is compared in blow-off side in the same u = mean velocity [m/s]
way as turbulent kinetic energy and its dissipation
rate. In the measurement result, it shows the References
maximum value at X= 25cm to 35cm. On the other [1] F. H. Rohles, S. A. Konz and B. W. Jones.
hand, in the CFD result, its value is smaller than the Ceiling Fans as Extenders of the Summer Comfort
measured value as a whole, and it becomes large in Envelope. ASHRAE Transactions, Vol.89 (1983),
the distant position from the rotational center of the Part 1A, pp.245-262
ceiling fan. [2] S. Chandra. Fans to Reduce Cooling Costs in
the Southeast. Florida Solar Energy Centre FSEC-
7. Conclusions EN-13-85, Cape Canaveral, FL, 1985
Air velocity near a ceiling fan was measured. [3] R. Aynsley. Fan Size and Energy Efficiency.
Then the CFD simulations of the experimented International Journal of Ventilation, Volume 1
space were performed using the boundary Number1, 2002
conditions of the ceiling fan based on air velocity [4] P.V. Nielsen. Description of Supply Openings
measurement, and validity of the boundary in Numerical Models for Room Air Distribution.
conditions was examined by comparing the ASHRAE Transactions (1992): symposia.
measurement result with the CFD result. The [5] STREAM ver.5, Software Cradle co. Ltd.
conclusions can be drawn as follows: http://www.cradle.co.jp/eindex.htm
• The floor affects the airflow pattern greatly in the
distant position from the ceiling fan.
• The measurement value and the CFD value were
in good agreement about the suction airflow.
• Under the downward condition, air velocity near
the rotational center of the ceiling fan in the CFD
result was larger than that in the measurement
result. As a cause, it is possible that large
distribution resulting from the rotational cycle of
the ceiling fan is also included in turbulent
parameters given as boundary conditions and that
the division interval of space is not fully fine.
In order to account for the influence of the division
interval of space, CFD simulations with fine
division are required. It may not be appropriate to
input the value computed from the measurement
result into CFD simulation as it is, and then the
further examination is required about turbulent
parameters given to the boundary conditions of the
ceiling fan. Moreover, the size and position of the
boundary surface should be examined, too.
More research would be needed to investigate the
effect of the size of the computational grid and the
turbulence parameters on the accuracy of the
simulation with this airflow model.

View publication stats