Tunnelling and

Underground Space
Technology
incorporating Trenchless
Technology Research
Tunnelling and Underground Space Technology 22 (2007) 474–482
www.elsevier.com/locate/tust

CFD simulation and optimization of the ventilation for subway

side-platform
Feng-Dong Yuan *, Shi-Jun You
School of Environment Science and Technology, Tianjin University, No. 72, Weijin Road, Nankai District, Tianjin 300072, China

Received 27 May 2006; received in revised form 23 September 2006; accepted 19 October 2006
Available online 5 December 2006

Abstract

To obtain the velocity and temperature field of subway station and the optimized ventilation mode of subway side-platform station,
this paper takes the evaluation and optimization of the ventilation for subway side-platform station as main line, builds three dimen-
sional models of original and optimization design of the existed and rebuilt station. And using the two-equation turbulence model as
its physics model, the thesis makes computational fluid dynamics (CFD) simulation to subway side-platform station with the boundary
conditions collected for simulation computation through field measurement. It is found that the two-equation turbulence model can be
used to predict velocity field and temperature field at the station under some reasonable presumptions in the investigation and study. At
last, an optimization ventilation mode of subway side-platform station was put forward.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Ventilation; CFD simulation; k-e model; Subway; Side-platform

1. Introduction as vehicle emission dispersion, visibility, air velocity, etc.

(Li and Chow, 2003; Yau et al., 2003; Gehrke et al., 2003).
Computational fluid dynamics (CFD) software is com- Earlier CFD simulations of tunnel ventilation system
monly used to simulate fluid flows, particularly in complex mainly focus on emergency situation as fire condition
environments (Chow and Li, 1999; Zhang et al., 2006; (Modic, 2003; Carvel et al., 2001; Casale, 2003). Many sci-
Moureh and Flick, 2003). CFD is capable of simulating a entists and research workers (Waterson and Lavedrine,
wide variety of fluid problems (Gan and Riffat, 2004; 2003; Sigl and Rieker, 2000; Gao et al., 2004; Tajadura
Somarathne et al., 2005; Papakonstantinou et al., 2000; et al., 2006) have done much work on this. This paper stud-
Karimipanah and Awbi, 2002). CFD models can be realis- ied the performance of CFD simulation on subway envi-
tically modeled without investing in more costly experi- ronment control system which has not been studied by
mental method (Betta et al., 2004; Allocca et al., 2003; other paper or research report. It is essential to calculate
Moureh and Flick, 2003). So CFD is now a popular design and simulate the different designs before the construction
tool for engineers from different disciplines for pursuing an begins, since the investment in subway’s construction is
optimum design due to the high cost, complexity, and lim- huge and the subway should run up for a few decade years.
ited information obtained from experimental methods (Li The ventilation of subway is crucial that the passengers
and Chow, 2003; Vardy et al., 2003; Katolidoy and Jicha, should have fresh and high quality air (Lowndes et al.,
2003). Tunnel ventilation system design can be developed 2004; Luo and Roux, 2004). Then if emergency occurred
in depth from the predictions of various parameters, such that the well-designed ventilation system can save many
people’s life and belongings (Chow and Li, 1999; Modic,
*
Corresponding author. Tel.: +86 22 8740 1917; fax: +86 22 2789 2626. 2003; Carvel et al., 2001). The characteristics of emergency
E-mail address: [email protected] (F.-D. Yuan). situation have been well investigated, but there have been

0886-7798/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tust.2006.10.004
 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482 475

few studies in air distribution of side-platform in normal 18.7 m(W) · 4.4 m(H), which is a typical side-platform
conditions. station. Each side has only one passageway (length ·
The development of large capacity and high speed com- height = 6.4 m(L) · 2.9 m(H)). The middle of station is
puter and computational fluid dynamics technology makes the space for passengers to wait for the vehicle. The plat-
it possible to use CFD technology to predict the air distri- form mechanical ventilation is realized with two jet open-
bution and optimize the design project of subway ventila- ings located at each end of station and the supply air jets
tion system. Based on the human-oriented design towards train and track. There is no mechanical exhaust
intention in subway ventilation system, this study simu- system at the station and air is removed mechanically by
lated and analyzed the ventilation system of existent station tunnel fans and naturally by the exits of the station.
and original design of rebuilt stations of Tianjin subway in
China with the professional software AIRPAK, and then 2.2. The design structure and ventilation of rebuilt station
found the optimum ventilation project for the ventilation
and structure of rebuilt stations. The predicted passenger flow volume increase greatly
and the dimension of the original station is too small, so
2. Ventilation system in the rebuilding design, the structure of subway station
is changed to, (length · width · height = 132 m(L) ·
Tianjin Metro, the secondly-built subway in China, will 17.438 m(W) · 4.65 m(H)), and each side has two passage-
be rebuilt to meet the demand of urban development and ways. The design volume flow of Southwest Station is
expected to be available for Beijing 2008 Olympic Games. 400 000 m3/h. For most existent stations, the platform
The existent subway has eight stations, with a total length height is only 2.9 m, which is too low to set ceiling ducts.
of 7.335 km and a 0.972 km average interval. For sake of So in the original design, there are two grille vents at each
saving the cost of engineering, the existent subway will con- end of the platform to supply fresh air along the platform
tinue to run and the stations will be rebuilt in the rebuilding length direction and two grille vents to jet air breadthways
Line 1 of Tianjin subway. Although different existent sta- towards trains. The design velocity of each lengthways
tions of Tianjin Metro have different structures and geom- grille vent is 5.54 m/s. For each breadthways vent, it is
etries, the Southwest Station is the most typical one. So the 5.28 m/s. Under the platform, 80 grille vents of the same
Southwest Station model was used to simulate and analyze velocity (4.62 m/s, 40 for each platform of the station)
in the study. Its geometry model is shown in Fig. 1. are responsible for exhaust.

2.1. The structure and original ventilation mode of existent 3. CFD simulation and optimization
station
The application of CFD simulation in the indoor envi-
The subway has two run-lines. The structure of South- ronment is based on conversation equations of energy,
west Station is, length · width · height = 74.4 m(L) · mass and momentum of incompressible air. The study

Fig. 1. The location of test section and the layout of measuring points.
476 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482

adopted a turbulence energy model that is the two-equa- sumption, the results of the field measurements at
tion turbulence model advanced by Launder and Spalding. the existent station can be used as velocity boundary
And it integrated the governing equation on the capital conditions to predict velocity filed of new station.
control volumes and discretized in the definite grids, at last
simulated and computed with the AIRPAK software.
3.2. Original conditions
3.1. Preceding simplifications and presumptions
To obtain the boundary conditions for computation and
Because of mechanical ventilation and the existence of simulation, such as the air velocity and temperature of
train-driven piston wind, the turbulence on platform is enclosure, measures were done by times at Southwest Sta-
transient and complex. Unless some simplifications and tion. All data are recorded during a complete pulling-in
presumptions are made, the mathematics model of three- cycle. The air velocities were measured by the multichannel
dimensional flow is not expressed and the result is diver- anemonmaster hotwire anemoscope and infrared ther-
gent. While ensuring the reliability of the computation mometer is used to measure the temperature of the walls
results, some preceding simplifications and presumptions of the station which are taken as the constant temperature
have to be taken. thermal conditions in the simulation.

(1) The period of maximum air velocity is paid attention 3.2.1. Temperatures of enclosure
to in the transient process. Apparently the maximum Divide the platform into five segments and select some
air velocity is reached at the period when train stops typical test positions. The distributing temperature of
at or starts away from the station (Yau et al., 2003; enclosure is shown in Table 1. It can be seen from Table
Gehrke et al., 2003), so the period the simulation con- 1 that all temperatures of enclosure are between 23 C
cerns about the best period of time for simulation is and 25 C, there is little difference in all test positions,
from the point when at the section of ‘x = 0.0 m’ and the average temperature is 24 C. So all temperatures
(Fig. 1) and the air velocity begin to change under of subway station’s walls is 24 C in CFD computation
piston-effect to the point when train totally stops at and simulation.
the station (defined as a ‘pulling-in cycle’).
(2) Though the pulling-in cycle is a transient process, it is 3.2.2. Time-averaged air velocity above the platform
simplified to a steady process. Fig. 1 is the location of test section and the layout of
(3) Because the process is presumed to a steady process, measuring points. The data measured include 12 transient
the transient velocity of test sections, which was velocities in each section (A–H in Fig. 1), which were deal
tested in Southwest Station in pulling-in cycle, is pre- with section’s time-averaged velocities in the period, 12
sumed to the time-averaged velocity of test sections. point’s velocities of passageway, which is used to acquire
(4) The volume flow driven into the station by pulling-in the average flow, and the velocities of each end of station,
train is determined by such factors as BR (blocking which is used to acquire the average piston flow volume.
ratio, the ratio of train cross-section area to tunnel Fig. 2 is the lengthways velocities measured of platform
cross-section area), the length of the train and the sections, Vmax is the maximum air velocity, Vmin is the
resistance of station etc. For existent and new sta- minimum air velocity and Vave is the average air velocity.
tions, BRs are almost the same. Although the length Fig. 2 shows that the maximum air velocity is at the pas-
of the latter train doubles that of the former which sageway. At the passageway the change of air velocity is
may increase the piston flow volume, the resistance about 2.25 m/s, which is the maximum and indicates that
of latter is greater than that of the former which the passageway is the position effected most by the piston
may counteract this increase. So it is presumed that wind effect, and the air velocity of section D and E after
the piston flow volume is same for both existent the passageway is almost the same, which indicates that
and new station and that the volume flow through the piston wind can hardly effect the air velocity after the
the passenger exits is also same. Based on this pre- passageway.

Table 1
Distributing temperature of enclosure
Test positions Station walls Platform
Top Railway ground Lateral wall Ground
1 2 3 1 2 3 1 2 3 1 2 3
Temperature (C) when x = 12.8 m 24.8 24.6 24.8 23.8 23.8 23.8 24.2 23.8 23.6 23.8 23.8 23.8
Temperature (C) when x = 25.6 m 24.8 24.6 24.6 24 24 24 24.4 24 23.6 23.4 23.8 24
Temperature (C) when x = 38.4 m 24.8 25 24.8 24 24 24 24.4 23.8 23.6 23.6 23.2 23.6
Temperature (C) when x = 51.2 m 24.4 24.4 24.4 24.2 24.2 24.2 24.2 24.2 23.4 23 23.4 23.4
 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482 477

station in CFD simulation is shown in Table 2. Addition-

ally, the Tianjin Metro is out of run during the measuring
time, so the heat load is not typical and is difficult to be
determined so that the distributing temperature was not
simulated in this study.
Referred to the feasibility research report of No. 3
design institute of China’s ministry of railway and the mea-
sured data in Southwest Station, the boundary conditions
of new station in CFD simulation of the existent station
is shown in Table 3.

3.3. Solution process

3.3.1. Method of solution

The governing equations to calculate original variables,
such as velocity and temperature etc., can be shown
(Moureh and Flick, 2003).
Fig. 2. Measured lengthways velocities at platform sections. o
ðq/Þ þ divðq~
u/Þ ¼ divðCgrad/Þ þ S ð1Þ
ot
In addition, because of the movement of train, the veloc- where / is common variable and denotes (u, v, w, T), ~ u is
ity cannot be measured in section U of Fig. 1, the air veloc- velocity vector, q is density, C is dissipation function, S is
ity of section U was computed in CFD simulation. the source item. u is air velocity of direction X, v and w
are air velocities of direction Y and Z, respectively, and
3.2.3. Boundary conditions of simulation T is air temperature.
Through analyzing the boundary data measured at the Eq. (1) is integrated on the spatial control volumes, and
Southwest Station, the boundary conditions of the existent through discretizing in the definite grids

Table 2
Boundary condition of existent stations in CFD simulation
Sections Section G Section U Section H Passageway of the train- Passageway of the non-train The exit of
(estimated) stopping side:EXIT1 stopping side:EXIT2 train F
Air velocity (m/s) 0.67 0.67 0.67 0.67 0.67 0.67
Air flow (m3/s) 0.67 0.67 0.67 0.67 0.67 0.67

Table 3
Boundary condition of rebuilt stations in CFD simulation
(a) Heat load
Persons Lamps (on the top Advertisement lamps (lateral Heat emitted from train
of platform) wall of platform)
Heat 44 kW/ 13 W/m2 15 kW/station 320 kW (acceleration) and 200 kW
load person (heat from applying brake)

(b) Air velocity

Supply air (lengthways) Supply air (breadthways) Under the platform vents
Air velocity (m/s) 5.54 5.28 4.62
(c) Boundary conditions of air velocity
The entrance of The exit of Passageway of the train-stopping Passageway of the non-train-stopping
train train side side
Boundary conditions 0.2–0.59 0.41 1.44 0.97
of air velocity (m/s)
(d) Temperature
Train walls (expert for excluding top Outdoor air Supply air Tunnel Enclosure
and bottom of train) (summer) (summer) air walls
Temperature 27 27 27 25 24
(C)
478 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482

Table 4
Settlement of existent station ventilation CFD simulation
Variable Pressure Kinetic k e
Discretization Body force weighted Second order upwind Second order upwind Second order upwind
Relaxation coefficient 0.3–1.0 0.3–1.0 0.3–1.0 0.3–1.0
Solution Pressure AMGa Kinetic AMG k AMG e AMG
Cycle type V-type Flexible Flexible Flexible
a
AMG is algebra multi-grid.

Table 5
Settlement of new station ventilation CFD simulation
Variable Pressure Temperature Kinetic k e
Discretization Body force weighted Second order upwind First order upwind First order upwind First order upwind
Relaxation coefficient 0.3–1.0 0.3–1.0 0.3–1.0 0.3–1.0 0.3–1.0
Solution Pressure AMG Temperature AMG Kinetic AMG k AMG e AMG
Cycle type V-type Flexible Flexible Flexible Flexible
X
ap / ¼ anb /nb þ b ð2Þ E is not equal to the values of the field measurement, whose
nb error is about 0.1–0.2 m/s, the differences are acceptable
In Eq. (2), nb denotes neighbor grids; ap and anb are the and the overall variance trend of the velocity along the
coefficients of / and /nb, respectively, b is the source item. platform tallies with the measurement. The simulation
(2) Tables 4 and 5 shows solution, cycle type and result of air velocity of 1.7 m height above platform of exis-
convergence criterion and relaxation coefficient in CFD tent station can be seen in Fig. 4. Fig. 4 shows that the
simulation of the existent station ventilation and the new majority of piston winds pour out from the passageway
station. and the air velocity of the zone after the passageway is
lower, which is below 0.4 m/s. The maximum air velocity
3.4. Simulation and discussion at 1.7 m height above platform is the passageway of the
train-stopping side, which is about 1.89 m/s.
Used the boundary conditions measured in Southwest The simulation results indicate that it is applicable to use
Station, the air velocity of subway side-platform station two-equation turbulence model to predict time-averaged
was simulated in AIRPAK. Fig. 3 is the air velocity com- flow and temperature distribution at subway station so
parison of simulation and measurement in section A–E, the veracity is reliable.
which shows that the two-equation model and the preced- Fig. 5 is the simulation result, which is the original
ing simplifications are acceptable in predicting flow at the design of air velocity at 1.7 m height above platform of
station. Although the computed velocity at section C and new station. It is shown from Fig. 5 that at the train-
ing-stopping side of platform the supply air does not
reach the zone between the two exits, which is for the
passenger to wait for the train, and pour out from the
passageway because of the leakage of passageway. The
air velocity between two passageways, which is below
0.4 m/s, is a little low so that the ‘Dead Zone’ is formed.
The simulation also shows that the maximum velocity
exists at the exits, which agrees with the references and
experience.
Fig. 6 is the temperature distribution at 1.7 m height
above platform of the new station, which is used to simu-
late for original design ventilation mode. It shows that
the temperature distribution is imbalanced for the leakage
of exits at the training-stop side of platform, and at each
end of platform, the temperature should be approximately
between 24 C and 26 C while the temperature is between
32 C and 34 C because the heat away from the train can
not be cooled and removed by the cold supply air in time,
even at the x = 61 m point, the temperature reaches above
40 C, which makes people feel uncomfortable.
The simulation result of temperature and air velocity of
Fig. 3. Air velocity of section A–E simulation and measurement. original design at the lengthways center section is shown in
 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482 479

Fig. 4. Simulation result of air velocity at 1.7 m height above platform of existent station.

Fig. 5. Simulation result of air velocity at 1.7 m height above platform of new station (original design).

Fig. 6. Simulation result of temperature at 1.7 m height above platform of new station (original design).

Fig. 7, which shows that the distribution of temperature (1) The ventilation mode that the feasibility research
and air velocity is not balanceable and the zone between report design for the side-platform station of subway
two exits is ‘Dead Zone’. makes the distribution of temperature and air veloc-
From the simulation of original design, the following ity unbalanced, so the passengers will feel uncomfort-
can be seen: able at the middle of platform.
480 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482

Fig. 7. Simulation result of temperature and air velocity at the lengthways center section (original design).

(2) The air velocity is too high at some position of plat- with original design’s CFD simulation. The simulation
form and the passengers feel easily the change of air results are showed in Figs. 8 and 9.
velocity, whose scope is between 0.4 m/s and 2.5 m/s, Fig. 8 is the air velocity distribution at 1.7 m height
and feel uncomfortable when they go to exits from above platform of new station, which is simulated for
the middle of platform. optimization design ventilation mode. It shows that the
(3) The efficiency of ventilation is not as good as expec- distribution of air velocity is good enough and balanced.
tation and the supply air from jet fans cannot reach The breadthways supplied air makes the ‘Dead Zone’
the middle of platform adequately and part of them disappeared, and the maximum air velocity is only
will flow away from the passageways. about 1.67 m/s at 1.7 m height above platform of new
station, which makes it difficult for passengers to feel
Because of the unbalanced distribution of temperature windy.
and air velocity, original design’s ventilation mode was Fig. 9 is the simulation result of temperature and air
optimized. The exhaust vents are as same as that of original velocity of original design at the lengthways center section,
ones. The flow of supplied air remains the same and under which shows that the ‘Dead Zone’ disappears at the length-
the platform there are tubes of supplied fresh air. Also ways center section. And the breadthways supplied air pre-
there are some vertical tubes fetched out by platform. As vents the heat, which is from the air conditioning
a result, the air is supplied widthways and reasonably the condenser and the friction produced by train-stopping,
flow of supplied air is delivered in balance, which elimi- from diffusing to the platform, so the temperature is about
nates the unbalanced distribution of temperature and air below 28 C.
velocity on the platform. On the other hand the breadth- Compared with the ventilation mode in original design,
ways supplied air prevents the heat brought by train from the optimization ventilation mode has several advantages
diffusing onto the platform, and the heat can be vented in as follows:
time from the exhaust vents under the platform. At the
same time the fresh air from the breadthways vents passes (1) The supply air is delivered to every zone in a balanced
through the passengers, takes away heat and humidity, and way and the distribution of temperature and air
the efficiency of ventilation is quite satisfactory. velocity of side-platform is also balanced.
To validate the optimization ventilation mode, the sim- (2) The air velocity is balanced and the maximum is only
ulation adopted the same boundary conditions and set up about 1.67 m/s.

Fig. 8. Simulation result of air velocity at 1.7 m height above platform of new station (optimization design).
 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482 481

Fig. 9. Simulation result of temperature and air velocity at the lengthways center section (optimization design).

(3) The supplied air removes the heat of platform better References
and the whole platform’s temperature is lower than
that of the original design, which can lower tempera- Allocca, C., Chen, Q.Y., Glicksman, L.R., 2003. Design analysis of single-
ture standard and can decrease the flow of supply air sided natural ventilation. Energy and Buildings 35, 785–795.
Betta, V., Cascetta, F., Labruna, P., Palombo, A., 2004. A numerical
to save energy. approach for air velocity predictions in front of exhaust flange slot
(4) The supplied air from breadthways vents also well openings. Building and Environment 39, 9–18.
prevents the hot contaminated train-driven air from Carvel, R.O., Beard, A.N., Jowitt, P.W., 2001. The influence of longitu-
spreading onto the platform, and on the other hand dinal ventilation systems on fires in tunnels. Tunnelling and Under-
the fresh air from the breadthways vents passes ground Space Technology 16, 3–21.
Casale, E., 2003. The automation of the aeraulic response in the case of a
through the passengers, takes away heat and humid- fire in a tunnel-first concrete answers. In: Proceedings of Claiming the
ity, collects and is exhausted by the under-platform Underground Space, pp. 185–191.
vents. Chow, W.K., Li, J.S.M., 1999. Safety requirement and regulations reviews
(5) The ventilation efficiency is quite satisfactory. on ventilation and fire for tunnels in the Hong Kong Special
Administration Region. Tunnelling and Underground Space Technol-
ogy 14, 13–21.
Gan, G.H., Riffat, S.B., 2004. CFD modelling of air flow and thermal
4. Conclusions performance of an atrium integrated with photovoltaics. Building and
Environment 39, 735–748.
The following can be concluded from this study: Gao, P.Z., Liu, S.L., Chow, W.K., Fong, N.K., 2004. Large eddy
The air flow in platform of subway is a complex three- simulations for studying tunnel smoke ventilation. Tunnelling and
Underground Space Technology 19, 577–586.
dimensional transient turbulence. Simplification of the air Gehrke, P.J., Stacev, C.H.B., Agnew, N.D., 2003. Rail tunnel temperature
flow to steady process and presumption of the transient stratification and implications for train and tunnel ventilation design.
velocity to the time-averaged velocity are applicable to sim- In: Proceedings of 11th International Symposium on Aerodynamics
ulate the distribution of temperature and air velocity of and Ventilation of Vehicle Tunnels, pp. 727–741.
Karimipanah, T., Awbi, H.B., 2002. Theoretical and experimental
subway platform in the pulling-in cycle.
investigation of impinging jet ventilation and comparison with
It is applicable to use two-equation turbulence model to wall displacement ventilation. Building and Environment 37, 1329–
predict time-averaged velocity and temperature distribu- 1342.
tion at subway station. The simulation results are influ- Katolidoy, J., Jicha, M., 2003. Eulerian-Lagrangian model for traffic
enced greatly by the boundary conditions such as dynamics and its impact on operational ventilation of road tunnel. In:
boundary velocity and temperature. Proceedings of 11th International Symposium on Aerodynamics and
Ventilation of Vehicle Tunnels, pp. 877–891.
The optimization ventilation mode has more advantages Li, J.S.M., Chow, W.K., 2003. Numerical studies on performance
than the original design’s. The breadthways ventilation is evaluation of tunnel ventilation safety systems. Tunnelling and
suggested when the height of platform is too low to have Underground Space Technology 18, 435–452.
ceiling ducts. It makes the distribution of temperature Lowndes, Ian S., Crossley, Amanda J., Yang, Zhi Yuan, 2004. The
and air velocity balanced. ventilation and climate modeling of rapid development tunnel
drivages. Tunnelling and Underground Space Technology 19, 139–150.
Before a design is accepted it is very necessary to evalu- Luo, S., Roux, B., 2004. Modeling of the HESCO nozzle diffuser used in
ate and optimize it by such techniques as CFD, etc. It will IEA Annex-20 experiment test room. Building and Environment 39,
help a lot to build a reliable, effective and economical 367–384.
system. Modic, J., 2003. Fire simulation in road tunnels. Tunnelling and
Underground Space Technology 18, 525–530.
Moureh, J., Flick, D., 2003. Wall air-jet characteristics and airflow
Acknowledgement patterns within a slot ventilated enclosure. International Journal of
Thermal Sciences 42, 703–711.
Papakonstantinou, K.A., Kiranoudis, C.T., Markatos, N.C., 2000.
This project was supported by the funding generated by
Computational analysis of thermal comfort: the case of the archae-
the Tianjin Municipal Science and Technology Commis- ological museum of Athens. Applied Mathematical Modelling 24, 477–
sion of China (Project No: 033112911). 494.
482 F.-D. Yuan, S.-J. You / Tunnelling and Underground Space Technology 22 (2007) 474–482

Sigl, O., Rieker, K., 2000. NATM tunnelling in Singapore old alluvium- Waterson, N.P., Lavedrine, J., 2003. Evaluation of a simplified model for
design assumptions and construction experience. In: International the design of ventilation systems in tunnel networks. In: Proceedings of
Conference on Tunnels and Underground Structures. pp. 205–212. 11th International Symposium on Aerodynamics and Ventilation of
Somarathne, S., Seymour, M., Kolokotroni, M., 2005. Dynamic thermal Vehicle Tunnels, pp. 219–229.
CFD simulation of a typical office by efficient transient solution Yau, R., Cheng, V., Lau, K.P., Lai, F., 2003. Design and analysis of
methods. Building and Environment 40, 887–896. tunnel ventilation system for the extension of an existing railway
Tajadura, R.B., Carlos, S.M., Eduardo, B.M., 2006. Influence of the slope tunnel in Hong Kong. In: Proceedings of 11th International Sympo-
in the ventilation semi-transversal system of an urban tunnel. sium on Aerodynamics and Ventilation of Vehicle Tunnels, pp. 323–
Tunnelling and Underground Space Technology 21, 21–28. 338.
Vardy, A.E., Mori, E., Yokota, M., Nakahori, I., 2003. Model-based Zhang, Lin, Feng, Jiang, Chow, T.T., Tsang, C.F., Lu, W.Z., 2006. CFD
predictive control of road tunnel portal emissions. In: Proceedings of analysis of ventilation effectiveness in a public transport interchange.
Claiming the Underground Space, pp. 207–212. Building and Environment 41, 254–261.