-
ARTICLE IN PRESS
Journal of Wind Engineering
and Industrial Aerodynamics 96 (2008) 2042–2053
0167-6105/$ -
doi:10.1016/j
�CorrespoE-mail ad
www.elsevier.com/locate/jweia
Wind load simulation for high-speed train stations
Nahmkeon Hura,�, Sa Ryang Kimb,Chan-Shik Wona, Chang-koon
Choic
aDepartment of Mechanical Engineering, Sogang University, Sinsoo
1, Mapo, Seoul 121-742, Republic of KoreabDepartment of Precision
Mechanical Engineering, Kangnung Nat’l. University, Jibyun 123,
Gangneung,
Gangwon 210-702, Republic of KoreacDepartment of Civil and
Environment Engineering, KAIST, Guseong 373-1, Yuseong, Daejeon
305-701,
Republic of Korea
Available online 7 April 2008
Abstract
A numerical simulation approach of wind load on buildings and
wind tunnel experiment are
presented in the present paper. Four Korean high-speed train
(KTX) stations (Cheonan-Asan,
Daejeon, Gwangmyeong and Gyeongju) were selected for this
purpose. For the numerical
simulation, 3-D incompressible flow with the standard k–e
turbulence model was adopted and thecommercial CFD software STAR-CD
was used. In order to validate results of the numerical
solution, a wind tunnel experiment was performed. Results from
the wind tunnel experiment using a
scaled model and CFD of four high-speed train station buildings
were compared, and good
agreement was achieved for the wind loads on the station
buildings. Hence, it was shown that CFD is
a good tool for predicting the wind load of huge and irregularly
shaped buildings.
r 2008 Elsevier Ltd. All rights reserved.
Keywords: Wind load; High-speed train station; Building;
Pressure; Computational fluid dynamics (CFD)
1. Introduction
To design new building structures, wind load acting on
structures should be consideredfor safety reasons. Wind load data
published for simple structures have been widely usedfor this
purpose (ASCE, 1998; Tamura et al., 1999). However, it is difficult
to use the data
see front matter r 2008 Elsevier Ltd. All rights reserved.
.jweia.2008.02.046
nding author. Tel.: +82 2 7058637; fax: +822 7120799.
dress: [email protected] (N. Hur).
www.elsevier.com/locate/jweiadx.doi.org/10.1016/j.jweia.2008.02.046mailto:[email protected]
-
ARTICLE IN PRESSN. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96
(2008) 2042–2053 2043
directly in relation to the complex structures. High-speed train
station buildings such asCheonan-Asan station, Daejeon station,
Gwangmyeong station, and Gyeongju station inKorea are huge in size
and have unique shapes and building structures. To consider the
safetyof these buildings under unusual weather conditions, wind
load should be obtained throughwind tunnel tests or computer
simulations, and taken into account in the design process.Computer
hardware and simulation techniques have been developing very
quickly. Therefore,it can be said that the computer simulation can
replace the expense and time spent conductingwind tunnel tests.
Recently, computer simulations have come into use for verifying
design inmany fields of architecture, mechanical engineering,
equipment, and the construction industry.
Gosman (1999) reviewed the capabilities and limitations of CFD
as a tool for windengineering, with particular reference to the
commercial CFD codes. He concluded that,although there are
well-known weaknesses in the physics modeling, the level of
predictionaccuracy was already sufficient for some purpose. Also,
for the same purpose, Huang et al.(2007) studied wind effects on
the Commonwealth Advisory Aeronautical Council(CAARC) standard tall
building with techniques of CFD, such as large eddy simulation(LES)
and the Reynolds-averaged Navier–Stokes equations (RANS) model,
etc. And thecomputed results were compared with experimental data
of wind tunnel. From the result,they showed that CFD techniques and
associated numerical treatments provided aneffective way for
designers to assess wind effects on a tall building.
For the curved–roofed building, which is increasingly used in
the modern builtenvironment, Blackmore and Tsokri (2006) performed
parametric wind tunnel studies andproperly scaled atmospheric
boundary layer simulation and gave an alternative to theEurocode
for wind actions (EN1991-1-4) recommended procedure. Ishii et al.
(2005)numerically calculated the actual house shapes. They showed
that the quality of predictionof wall surface wind pressure
distribution of various buildings was improved by applyingRNG and
Durbin turbulent models.
In this paper, results will be introduced from the wind tunnel
test and computersimulation on the effect of the wind load
generated from high-velocity winds like typhoons,for the station
buildings mentioned above. The wind tunnel test was carried out
atPOSTECH, and the computer simulation was performed with a
commercial CFD code,STAR-CD (CD-Adapco, 2006).
2. Wind tunnel test
POSTECH’s subsonic wind tunnel (POSWIT, Fig. 1) was used to
measure the wind loadon the scaled station models. The construction
of the POSWIT began in 1993 and wasinaugurated in November 1995.
The POSWIT provides a large-sized test section and highflow quality
with a large contraction ratio (9:1). The wind pressure was
measured with aPSI8400 electric pressure scanning system for 320
pressure taps. The measurement range ofpressure was about 5000 Pa
and the accuracy of the system was within 0.05% (2.5 Pa) offull
scale. The scan rate reached 20,000 readings/s. The specifications
of the wind tunnel aresummarized in Table 1, and Fig. 2 shows the
schematic diagram of the measurementequipment mounted in the wind
tunnel and dimension of the systems. The tested modelshave a 1/400
reduced scale and 320 pressure taps of which the inside hole
diameter is 1mm.The turntable under the station model controls the
wind yaw angle and the Pitot tubeabove the model measures the
reference static pressure. The measured data are recordedwithin 10
s and the averaged pressure coefficient is stored in the computer.
Generally,
-
ARTICLE IN PRESS
Fig. 1. POSTECH subsonic wind tunnel (POSWIT).
Table 1
The specifications of the wind tunnel
Type Closed/open type circuit wind tunnel
Size of test section 1.8 W� 1.5 H� 4.3 L (m), closed typeOverall
dimension of wind tunnel 14 W� 6 H� 37 L (m)Maximum wind velocity
75m/s
Minimum wind velocity 5m/s
Contraction ratio 9:1
Flow uniformity 0.25%
Turbulent intensity 0.2%
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008)
2042–20532044
dynamic viscosity of the air in the wind tunnel is equal to that
of natural wind, thus thesimilarity is satisfied when the product
of characteristic length and flow velocity is the samebetween the
scaled model and the real structure. In the present study, the
model is scaledby 1/400; hence, the wind velocity of the wind
tunnel is 400 times faster than in a realsituation. However, it is
well-known that the pressure coefficients are velocity
independentwith turbulent flow. Therefore, we compared the pressure
coefficient between the scaledand the computational model.
3. Numerical simulation
The commercial CFD code STAR-CD is used to solve the continuity
Equation (1) and3-D RANS Equation (2). To express the realizable
turbulence, a high Reynolds number
-
ARTICLE IN PRESS
Fig. 2. Schematic diagram of measurement equipment mounted in
the wind tunnel.
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008) 2042–2053
2045
k–e model (3), (4) was selected. Of course, the standard k–e
model has disadvantages suchas the overestimation of wind pressure
coefficient and turbulent kinetic energy on thewindward surface.
However, it is known that the k–e model is practical to use
generally,because this method is inexpensive and is used widely for
complex building models:
qrqtþ q
qxjðrujÞ ¼ sm (1)
qruiqtþ q
qxjðrujui � tijÞ ¼ �
qpqxiþ si (2)
qqtðrkÞ þ q
qxjrujk � mþ
mtsk
� �qkqxj
� �¼ mtðpþ pBÞ � r�
� 23
mtquiqxiþ rk
� �quiqxiþ mtPNL (3)
qqtðr�Þ þ q
qxjruj�� mþ
mts�
� �q�qxj
� �¼ C�1
�
kmtP�
2
3mt
quiqxiþ rk
� �quiqxi
� �þ C�3
�
kmtPB
� C�2r�2
kþ C�4r�
quiqxiþ C�1
�
kmtPNL (4)
where t is the time, xi is the Cartesian coordinate (i ¼ 1,2,3)
and ui is the absolute fluidvelocity component in direction xi.
Also, p is the piezometric pressure, r is the density, tij isthe
stress tensor components, and sm and si are the mass and momentum
source,respectively. In Eqs. (3 and 4), k is the turbulent energy,
e is the dissipation rate, and the mtis the turbulent viscosity.
Pressure–velocity coupling is taken care of by the SIMPLEalgorithm
(Patankar and Spalding, 1972) and pressure interpolation is second
order (Rhieand Chow, 1983). Second-order discretization schemes are
used for both the convectionterms and the viscous terms of the
governing equations. Numerical simulation was
-
ARTICLE IN PRESS
Table 2
Representative simulation cases
Case # Velocity of wind (m/s) Yaw angle (deg) Attack angle
(deg)
1 30 0 0
2 30 45 0
3 30 90 0
4 30 0 15
5 40 0 0
6 40 45 0
7 40 90 0
8 40 0 15
9 50 0 0
10 50 45 0
11 50 90 0
12 50 0 15
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008)
2042–20532046
calculated using the SGI ORIGINE machine (R10000CPUx4, 512MB
RAM, 18GB HDD,Cheonan-Asan, Daejeon, and Gwangmyeong station) and
Linux cluster (Dual Intel Xeon2.4GHz CPUs, 13 Nodes, 25GB RAM,
1.86TB HDD, Gyeongju station). Therepresentative simulation cases
are listed below in Table 2. As occasion demands, thecases can be
added. The ambient pressure boundary is assigned to the exterior
surface ofthe domain and the inlet velocity boundary is used to
represent the typhoon velocity. Thecomputation time takes 4 h using
1 node of the Linux cluster.
4. Results and discussion
4.1. Cheonan-Asan station
Cheonan-Asan station is the first station designed for
high-speed trains in Korea. Fig. 3shows the drawings of the station
buildings. The shape of the roofs looks like airfoils, andthey seem
unsafe under high wind speeds. First, the wind tunnel test for a
scaled model andthe computer simulations are undertaken to study
the effects of a typhoon using the above-described system.
Subsequently, the numerical simulation is performed under the
sameconditions. Fig. 4 presents the computational mesh for typhoon
effect simulation. About750,000 meshes are used for the simulation.
In the figure, the top-left side is the wholedomain, and the
lower-right side is the solid cell of the station.The comparisons
between the two results for the distributions of the pressure
coefficients
(5) at the two sections of the roofs (section B is the center of
the roof, section D is theposition between center and end of the
roof) are shown in Fig. 5:
Cp ¼p� p11=2 rU2
(5)
where Cp is the pressure coefficient, p the pressure, U the
velocity of wind, and subscriptNthe ambient. They show good
agreement in the magnitudes and profiles for the
pressurecoefficients. Thus, the result of the computer simulation
can be considered reliable. Fig. 6shows the velocity vector
distributions around the station and the roof when the windblows at
40m/s. The velocity over the roofs is faster than under the roofs.
Also, the
-
ARTICLE IN PRESS
Fig. 3. Drawing of Cheonan-Asan station.
Fig. 4. Computational domain of Cheonan-Asan station.
Fig. 5. Comparison of pressure distributions results between
wind tunnel test and CFD simulation.
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008) 2042–2053
2047
pressure distribution in Fig. 7 shows higher pressures on the
upper side of the roofs andlower pressures below the lower side.
Therefore, lift forces are expected to act upon theroofs.
-
ARTICLE IN PRESS
Fig. 6. Velocity profiles around the roof & station.
Fig. 7. Pressure distribution on the roofs.
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008)
2042–20532048
4.2. Daejeon station
The wind tunnel test of a scaled model and computer simulations
are carried out tostudy the effects of a typhoon on the Daejeon
station. The directions and magnitudes ofthe wind velocity for the
simulation were determined from climate data taken over the last50
years. For the safety of the passengers and the facilities in the
platform, the velocity andpressure data are obtained and analyzed.
Fig. 8 shows the 1/400-scaled model in the windtunnel, and the
drawings of the roofs of the station building are shown in Fig. 9.
Thepressure coefficient distributions on the roof from the wind
tunnel test are compared withthose of the computer simulations. The
positions of peak values show good agreement inFig. 10. In the
figure, the hollowed symbol (U) denotes the experimental results,
and thesolid symbol (Uni, ABL) represents the numerical values. In
addition, Uni and ABLdenote uniform velocity conditions and
atmospheric boundary layer conditions,respectively. From the
results, the computer simulation shows good results and
reliability.
4.3. Gwangmyeong station
The wind loads for the roofs and walls of the station buildings
are estimated by the windtunnel tests and the computer simulations,
and the results are compared. Fig. 11 shows the
-
ARTICLE IN PRESS
Fig. 8. Wind tunnel scaled model of Daejeon station.
Fig. 9. Various roof shapes of Daejeon station.
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008) 2042–2053
2049
cross-section drawing of Gwangmyeong station, and Fig. 12 shows
the computationalmeshes. As mentioned in Section 4.1, the top-left
side of the figure is the whole domain,and lower right is the solid
cell of the station.
Fig. 13 shows the scaled model for the wind tunnel tests. The
result from the wind tunneltest is compared with that from the
computer simulation in Fig. 14. The distributions ofthe pressure
coefficient show good agreement, thus the numerical results can be
used todesign the proper roof shape.
4.4. Gyeongju station
From the results of the above three cases of stations, the
numerical method isthoroughly validated. Therefore, for Gyeonju
station, only the numerical simulation isperformed. A total of 18
cases, including the effect for the various velocities and
directionsof the wind, are simulated by CFD. In this paper, only
one case is presented. Fig. 15 showsthe drawings for the station
building. Fig. 16 shows the computational meshes from the
-
ARTICLE IN PRESS
Fig. 10. Pressure distribution on roof 1 of Daejeon station.
Fig. 11. Cross-section drawing of Gwangmyeong station.
Fig. 12. Computational domain of Gwangmyeong station.
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008)
2042–20532050
-
ARTICLE IN PRESS
Fig. 13. Wind tunnel scaled model of Gwangmyeong station.
Fig. 14. Comparison of pressure distributions results between
wind tunnel test and CFD simulation.
Fig. 15. Drawing of Gyeongju station.
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008) 2042–2053
2051
-
ARTICLE IN PRESS
Fig. 16. Computational domain of Gyeongju station.
Fig. 17. Velocity vectors at the central cross-section of
Gyeongju station.
Fig. 18. Pressure contours at the central cross-section of
Gyeongju station.
N. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96 (2008)
2042–20532052
drawings. The total number of meshes is about 4.2 million solid
structures and the roofused 0.45 million meshes. The velocity
distributions around a central cross-section of thestation for the
35m/s of wind velocity are shown in Fig. 17. The velocities around
roofs arefast, but those around the platform are slow due to the
wind barrier. The pressuredistribution for the same case as shown
in the previous figure is depicted in Fig. 18. The
-
ARTICLE IN PRESSN. Hur et al. / J. Wind Eng. Ind. Aerodyn. 96
(2008) 2042–2053 2053
integration of the pressure around the roof gives the total wind
load acts on the roof. Thatvalue is used to design the structure of
the roofs and the station building.
5. Concluding remarks
In the present study, numerical simulations of wind load on
Korean high-speed train(KTX) stations are presented. The numerical
the results are compared to results from windtunnel tests, and
showed good agreement on wind load distribution. From the
presentstudy, it is believed that CFD can be successfully applied
to the prediction of wind loads onhuge and irregularly shaped
buildings, whose wind tunnel tests are expensive. Hence abroad
application of CFD on wind load is expected in the field of civil
and architecturalengineering.
Acknowledgments
This research was financially supported by the Korea High Speed
Rail ConstructionAuthority (Currently Korea Rail Network Authority)
and Ministry of Construction &Transportation of Korea. Authors
wish to thank the authorities for their kind support.
References
American Society of Civil Engineers, 1998. Minimum design loads
for buildings and other structures. ASCE
Standard No. 7-98.
Blackmore, P.A., Tsokri, E., 2006. Wind loads on curved roofs.
J. Wind Eng. Ind. Aerodyn. 94 (11), 833–844.
CD-Adapco, 2006. STAR-CD User Guide Version 3.26.
Gosman, A.D., 1999. Developments in CFD for industrial and
environmental applications in wind engineering.
J. Wind Eng. Ind. Aerodyn. 81 (1), 21–39.
Huang, S., Li, Q.S., Xu, S., 2007. Numerical evaluation of wind
effects on a tall steel building by CFD. J. Constr.
Steel Res. 63, 612–627.
Ishii, M., Kurabuchi, T., Endo, T., Maruta, E., Sawachi, T.,
2005. Study on the numerical predictive accuracy
of wind pressure distributions and air flow characteristics—Part
1. Optimization of turbulence models
for practical use; Part 2. Prediction accuracy of wind pressure
distribution of various shaped buildings.
Int. J. Vent. 4, 269–284.
Patankar, S.V., Spalding, D.B., 1972. A calculation procedure
for heat, mass and momentum transfer in three-
dimensional parabolic flows. Int. J. Heat Mass Transfer 15,
1787–1806.
Rhie, C.M., Chow, W.L., 1983. A numerical study of the turbulent
flow past an isolated airfoil with trailing edge
separation. AIAA J. 21, 1525–1532.
Tamura, K., Ohkuma, T., Okada, H., Kanda, J., 1999. Wind loading
standards and design criteria in Japan.
J. Wind Eng. Ind. Aerodyn. 83 (3), 555–566.
Wind load simulation for high-speed train
stationsIntroductionWind tunnel testNumerical simulationResults and
discussionCheonan-Asan stationDaejeon stationGwangmyeong
stationGyeongju station
Concluding remarksAcknowledgmentsReferences