A NUMERICAL ANALYSIS OF THE VENTILATION PERFORMANCE FOR DIFFERENT VENTILATION STRATEGIES IN A SUBWAY TUNNEL*

HUANG Yuan-dong

School of Environment and Architecture, University of Shanghai for Science and Technology, China

College of Engineering, Kyung Hee University, Yongin 449-701, Korea, E-mail: hyd1119@tom.com

LI Chan

Graduate School, Kyung Hee University, Yongin 449-701, Korea

KIM Chang Nyung

College of Engineering, Kyung Hee University, Yongin 449-701, Korea

Industrial Liaison Research Institute, Kyung Hee University, Yongin 449-701, Korea

A NUMERICAL ANALYSIS OF THE VENTILATION PERFORMANCE FOR DIFFERENT VENTILATION STRATEGIES IN A SUBWAY TUNNEL*

HUANG Yuan-dong

School of Environment and Architecture, University of Shanghai for Science and Technology, China

College of Engineering, Kyung Hee University, Yongin 449-701, Korea, E-mail: hyd1119@tom.com

LI Chan

Graduate School, Kyung Hee University, Yongin 449-701, Korea

KIM Chang Nyung

College of Engineering, Kyung Hee University, Yongin 449-701, Korea

Industrial Liaison Research Institute, Kyung Hee University, Yongin 449-701, Korea

An unsteady three-dimensional analysis of the ventilation performance is carried out for different ventilation strategies to find out a ventilation method with a high performance in a subway tunnel. The natural ventilation performance associated with a train-induced air flow in a subway tunnel is examined. The dynamic layering method is used to consider the moving boundary of a train in the current CFD method. The geometries of the modeled tunnel and the subway train are partially based on those of the Seoul subway. The effects of the structure of the ventilation duct and the geometry of the partitions on the ventilation performance are evaluated. The results show that the combined ventilation ducts (to be designed),and the partitioning blocks installed along the middle of tunnel (already in existences) are helpful for air exchange. This study can provide some guidance for the design of ventilation ducts in a subway system.

ventilation performance, subway tunnel, train-induced air flow, natural ventilation duct, numerical analysis

Introduction

With the rapid development of subway systems, people pay more and more attention to subway environmental aspects, including the temperature, humidity, air velocity, pressure, noise and so on. Roh et al.[1]performed a fire simulation and evacuation simulation to estimate the life safety in a subway fire. Fukuyo[2]studied the effect of task-ambient air-conditioning system on the air-conditioning loads of a subway station and the thermal comfort of passengers. It shows that this system can improve thermal comfort and decrease air-conditioning loads.

As an important issue in the subway environmental control system design, much work was conducted on the air quality problems in the subway system. Yuan and You[3]obtained the velocity and temperature field of a subway station and the optimized ventilation mode of the subway side-platform station by using CFD simulation. Ke et al.[4]explored the influence of various operations on the subway environment by combining the Subway Environmental Simulation Program with CFD software PHOENICS. Li et al.[5]studied the transient aerodynamic characteristics around vans running into a road tunnel. Kim and Kim[6]investigated the pressure and air velocity variations with time by conducting both experimental and numerical analyses. The predicted numerical model results see a good agreement with the experimental data. Kim and Kim[7]also found that the ventilation performance is influenced by the vent shaft location and that the optimum location of the vent shaft maximizing the ventilation performance should be near the station. Jia et al.[8]investigated the air flow characteristics in a subway station in various situations by utilizing CFD method. The results indicate that the pistoneffect has a significant influence on the flow field in the station and plays an important role in the natural ventilation. A two-dimensional CFD analysis of two trains passing by each other was performed by using a point symmetric method[9].

One of the main issues in developing a numerical method for simulating train-induced airflows in a subway tunnel is how to model the moving boundaries of the train (an immersed solid). There are two kinds of methods to model the moving boundaries of an immersed solid: the moving grid method and the fixed grid method.

The application of the moving grid method is closely related to the re-generation of a grid system. For the analysis of a flow field with a train motion in a tunnel, the chimera grid system, the patched grid system, and the adaptive remeshing grid system were used[4]. Also, the dynamic layering method was considered for the analysis of the air flow in a model subway tunnel[10]. As for the application of the fixed grid method, with the use of the sharp interface method[6,7,11]Kim and Kim[6]recently carried out a numerical analysis of the train-induced unsteady airflow inside their experimental model tunnel.

In a subway, the airflow in a ventilation duct is greatly influenced by the train-induced air flow in the tunnel. If a ventilation opening is located at an appropriate position, the natural ventilation can be accomplished by the piston effect caused by a moving train in the subway tunnel. Although there were studies of the underground tunnel ventilation associated with the piston effects caused by train motion[12], they are not applicable to the Seoul subway because their conditions such as the geometric dimensions of tunnels and subway train as well as the train run schedules are different from those in the Seoul subway. Especially, there were no quantitative analyses of the effects of the train-induced unsteady airflow in the tunnel space on the natural air ventilation in a real situation. Therefore, it is necessary to study unsteady three-dimensional airflows in a tunnel so as to provide some basic data for optimal designs of a tunnel ventilation system and for other environmental issues related to the traininduced flows (such as the evaluation of particle motion and particle distribution in a tunnel and in ventilation ducts, and the determination of appropriate positions for dust control filters in a tunnel).

The present work investigates numerically the ventilation performance of different ventilation strategies to obtain a high ventilation performance in a subway tunnel. Different geometries of the subway systems are considered using the dynamic layering method for the moving boundary of a train. The numerical results are analyzed to elucidate the ventilation performance for different ventilation strategies in association with a train motion in a natural ventilation environment.

1. Computational models

1.1 Problem description

The air flow in subway tunnels, stations (with platform screen doors) and ventilation ducts is a complex unsteady, three-dimensional turbulent flow with negligible compressibility. The modeled subway tunnel consists of two parallel tracks with each track of 4.1 m wide and 5.95 m high. The cross sections of one track and the train are illustrated in Fig.1. Two stations are located at the two ends of the tunnel. In this simulation, we consider the case where only a train is running on one track.

Fig.1 Cross section of a track and a subway train for a subway line (not in scale)

Fig.2 The geometry of the ventilation ducts on the tunnels (not in scale)

Two structures of the ventilation ducts are considered.

(1) Duct Type A: Three ventilation ducts with the identical geometry are installed on the ceiling of each track (see Fig.2(a)). The cross section of the ventilation duct is shown in Fig.3(a).

(2) Duct Type B: Two openings at the ceiling of the two tracks are connected to a ventilation duct (see Fig.2(b)). The cross section of the ventilation duct is shown in Fig.3(b).

For the purpose of comparison, apart from the case of no partition, a case with specially designed partitions is considered to see their influence on the ventilation performance during the train motion.

(1) Partition Type A: Without partition.

(2) Partition Type B: Three blocks (each of 150 m long) are installed near the ventilation ducts and four partitions with rectangular holes are used in the remaining parts (see Fig.4(b)).

Fig.3 Cross sections of two different designs of ventilation ducts (not in scale)

Fig.4 Geometries of the partition between the two tracks of Partition Type B (not in scale)

Fig.5 Side-view of the tunnel including station zones (not in scale)

Fig.6 Variation of train velocity with time

The length of the tunnel is 1140 m (see Fig.5) and the train (100 m long) moves 75 s with the velocity profile as depicted in Fig.6, which can be expressed as

where UTis the train velocity and t is the time.

Table 1 Geometries of the ventilation system

Numerical simulations are carried out for three cases of different geometries of the subway as shown in Table 1. The Case 1 was investigated by Huang et al.[13]

1.2 Governing equations

In the simulations, the air is considered incompressible. The air flow in the subway system is of a high Reynolds number turbulent flow. The continuity equation and the Reynolds-averaged Navier-Stokes equation for the turbulent flow are as follows:

where ν and νtare the laminar and turbulent kinematic viscosities of the air, respectively. The latter should be determined according to a suitable turbulence model. In the present study, the Standard k-ε turbulence model is used to determine the turbulent kinematic viscosity. The governing equations of the turbulent kinetic energy k and the dissipation rate of the turbulenceε are as follows:

The model constants are given by Launder and Spalding[14]as

1.3 Initial and boundary conditions

At t=0s, the pressure and the air velocity in the tunnel and the ducts are assumed to be zero, when the train stays in the left station. In the time domain of t＞0s, the pressure and the air velocity gradient at the ground openings of all ducts and at the left and right sides of the tunnel are zero.

2. Numerical procedure

A finite-volume method is adopted to discretize the three-dimensional governing equations by using the commercial software Fluent 6.3. The PISO algorithm is applied for the pressure-velocity coupling. The second-order upwind scheme is used to discretize the convection terms and the central difference scheme is employed for the diffusion terms. The PRESTO scheme is employed to solve the pressure corrective equation, which provides an improved pressure interpolation in cases of large pressure gradients. For unsteady analysis, the time derivatives are discretized with the first-order implicit scheme.

The cell size in the longitudinal direction Δx is chosen to be 0.6 m while the cell sizes Δy and Δz are 0.3 m and 0.4 m, respectively, except in places near the tunnel wall and in the narrow gap between the surface of the moving train and the tunnel wall where finer grids are used. The Duct Type A has around 1 892 000 hexahedral grids while the Duct Type B employs 1 489 000 grids.

Fig.7 Mass flow rates of the airflow in Case 2

The dynamic layering method is adopted for modeling the moving boundaries of the train. During the train motion the blocks of grids in front of the train are compressed while those in rear of the train are elongated. Thus, the layers of cells adjacent to the boundaries of the train can be removed or created, respectively, as the train moves. In this manner, the update of the volume meshes with the motion of a train ishandled at each time step, which is set to 0.01 s. The convergence criterion is that the normalized residual for each variable is less than 10?3.

3. Results

The current numerical method was validated by Huang et al.[13]against the numerical and experimental results in a modeled subway tunnel obtained by Kim and Kim[6].

The variations of the mass flow rates with time at the three ventilation ducts in Case 2 are shown in Fig.7(a). The positive and negative values of the mass flow rate denote the inflow and the outflow of the air through the ducts, respectively. As the train moves towards the ventilation Duct 1, the mass flow rate of the air outflow through Duct 1 increases with time because the pressure at the opening of Duct 1 increases and reaches its peak value at t=25s. As the train passes the ventilation Duct 1, the pattern of the outflow in the duct is changed very quickly into that of the inflow, and the mass flow rate of the air inflow through Duct 1 increases with time because the pressure at the opening of Duct 1 decreases drastically, and reaches its peak value when the rear face of the train passes the opening of the ventilation Duct 1 at t=30s. After the train moves away from the ventilation Duct 1, the mass flow rate of the air inflow through Duct 1 decreases with time because the pressure at the opening of Duct 1 is recovered with time. Generally, very similar flow patterns are observed in Duct 2 and Duct 3 in association with the train motion passing each opening of Duct 2 and Duct 3.

However, just after the pattern of the outflow in Duct 1 is changed into that of the inflow (approximately in the interval 26s＜t＜32s ), the exhaust mass flow rates (absolute value) at both Duct 2 and Duct 3 are increased. The increase in the mass flow rate in Duct 2 is more notable as compared with that in Duct 3 since the distance between Duct 1 and Duct 2 is smaller than that between Duct 1 and Duct 3. Just after the pattern of the outflow in Duct 2 is changed into that of the inflow (approximately in the interval of 36s＜t＜42s ), the mass flow rate of the suction flow through Duct 1 is reduced and that (absolute value) of the exhaust flow through Duct 3 is increased.

When the train almost reaches the right station, the exhaust airflows through Duct 2 (in the interval of 71.40s＜t＜75s ) and Duct 3 (in the interval of 66.40s＜t＜75s ) are observed, which are caused by the fact that the air in the front and the rear of the train moves together with the train, and when the train velocity is quite reduced, the air motion in the rear of the train is blocked by the train so that some part of the air is pushed out through the ducts. Since the distance between the right station and Duct 3 is smaller than that between the right station and Duct 2, the mass flow rate (absolute value) at Duct 3 is larger. interval of 0s＜t＜19.6s due to the presence of the left side where the gauge pressure is set to zero as the boundary condition. After t =19.6s, as the train moves away from the left side, the air comes into the left side of the tunnel towards the end.

The mass flow rate through the left side is shown in Fig.7(b). The incoming flow rate through the left side of the track where the train runs increases at first as the train accelerates. Just like a moving body in the air which induces the air motion near the surface of the body, especially, in an external flow, the train motion tends to induce the air flow in the other track in the reverse direction, which increases the outgoing mass flow rate on the left side of the other track in the

The mass flow rates through the right side are shown in Fig.7(c), where the outgoing flow rate (absolute value) in the interval of 0s＜t＜50s increases with time, the train accelerates and cruises with high speed. At about 57 s, the two mass flow rates separate suddenly and in the interval of 57s＜t＜61s the outgoing mass flow rate (absolute value) of the track where the train runs increases, while that of the other track decreases sharply. As the train approaches the right side, the train motion tends to induce the airflow in the other track due to the presence of the right side (where the gauge pressure is set to zero as the boundary condition). In the interval of 61s＜t＜75s, the outgoing mass flow rate (absolute value) of the track where the train runs decreases while that of the other track increases because the velocity of the train decreases, accompanied with decreased induction of the airflow in the other track.

Table 2 Mass flow of air through the three ventilation ducts and the left and right sides when 0s＜t＜75s in Case 2

The total mass flows of air through the three ventilation ducts and across the left and right sides in Case 2 are shown in Table 2. Here, the net mass flowin is 0.048 kg, which shows that the computational result is of a high accuracy.

Fig.8 Mass flow rates of the airflow in Case 3

Figure 8(a) illustrates the variations of the mass flow rate with time at the six ventilation ducts in Case 3. Three blocks are taken as partitions between the two tracks as shown in Fig.4. The patterns of the mass flow rates at Duct 1, Duct 2 and Duct 3 in Fig.8(a) are, in some sense, similar to those in Case 1, but the patterns at Duct 4, Duct 5 and Duct 6 are obviously different. When the train moves towards the block near Duct 1, the exhaust mass flow rates through Duct 1 and Duct 4 increase with time. Before the front face of the train reaches the leading edge of the first block near Duct 1, the air in front of the train can move easily to the other track. When the train reaches the leading edge of the first block near Duct 1 at t =21s, the mass flow rates of Duct 1 and Duct 4 begin to separate. Then, the mass flow rate of the air exhausted through Duct 1 sharply increases to its peak value since the air in front of the train cannot move to the other track, while the mass flow rate of the air pushed out through Duct 4 decreases and even the air suction through Duct 4 can be observed. When the exhaust air flux through Duct 1 is quite large in the interval of 21s＜t＜25s, slightly increased exhaust air fluxes through Ducts 2 and 5, and Ducts 3 and 6 are observed. When the train passes the opening of Duct 1, the exhaust air flow through Duct 1 is suddenly changed to the suction air flow at t=26s. The variation of mass flow rates in the other four ducts as shown in Fig.8(a) can be explained in a similar way.

The mass flow rates through the left side and the right side in Case 3 are shown in Figs.8(b) and 8(c). The patterns of the mass flow rate are similar to those shown in Figs.7(b) and 7(c) for Case 2 without partitions near the left and right sides, but the maximum mass flow rates are smaller than those in Case 2. The total mass flows of the air through the ventilation ducts and across the left and right sides in Case 3 are shown in Table 3.

Table 3 Mass flow of air through the six ventilation ducts and the left and right sides during 0s＜t＜75s in Case 3

4. Discussion

The patterns of mass flow rates for Case 2 are similar to those in Case 1 shown in Fig.9(a)[13]. One of the differences in the patterns between Case 1[13]andCase 2 is that the values of mass flow rate are generally larger in Case 2 than those in Case 1[13], due to the fact that in Case 2 two openings at the ceiling of the two tracks are connected to one ventilation duct so that the effective cross-sectional area of the ventilation ducts for the air to pass through in Case 2 is smaller than that in Case 1, resulting in a higher mass flow rate for each duct in Case 2. The patterns of timevarying mass flow rates in Case 2 are smoother than those in Case 1[13]. It may be caused by the fact that in Case 2 there are only three ventilation ducts with a smoothing effect for physical fluctuations at two openings for each duct, while in Case 1 there are six ducts which may generate complicated time-varying mass flow rates at different locations in association with time-varying pressures at the openings of different ducts.

Fig.9 Mass flow rates of the airflow in Case 1[13]

Big difference in mass flow rates at various ducts21s＜t＜33.75s (from the time when the front of the train enters the block region to the time when the rear of the train leaves the block region). In Case 3 (Fig.8(a)), because of the existence of Duct 1 (on the track where the train runs) and Duct 4 (on the other track) located in the middle of the blocked region (Fig.4(b)), the mass flow rate of the exhaust flow at Duct 1 is notable in the interval of21s＜t＜25.8s while the mass flow rate at duct 4 is not considerable during the above period. However, before t=21s, the patterns in both Case 1[13]and Case 3 are almost the same. After the rear of the train leaves the firstblock region, the mass flow rate at Duct 4 is quite smaller than that at Duct 1 as shown in Fig.8(a) because the blocks suppress the suction flow at Ducts 4, 5 and 6 with the train motion on the track (inclu-In Fig.9(b)[13], after t = 2 5 s ,rate of the air coming from the left side of the track where the train runs reduces gradually, while, on the other hand, that shown in Fig.7(b) increases. This may mean that in the whole subway tunnel the piston effect due to a train motion is more notable in Case 2 than in Case 1[13]because of the small effective cross sectional area of the ventilation ducts in Case 2. However, the patterns of the mass flow rates shown in Fig.8(b) and Fig.9(b)[13]are almost the same, which means that the geometry of the partition has little effect on the air flow through the left sides of the tracks.

The patterns shown in Fig.7(c) and Fig.9(c)[13]are almost the same before t=57s. The maximum mass flow rate of the right side of the track where the train runs in Case 2 is larger by 20 kg/s than that in Case 1[13]at t=62s, showing a notable piston effect in view of the smaller effective cross sectional area of the ducts in Case 2. The mass flow rate of the right side of the other track in Fig.7(c) is always smaller than 0 (exhaust flow), while the mass flow rate of the right side of the other track in Fig.9(c)[13]is larger than 0 (suction flow) around t=64s when the air inflow at the right side in the direction opposite to the train motion is observed because of the air flow around the running train.

The air flow patterns shown in Fig.8(c) and Fig.9(c)[13]are almost the same when t<75s. The maximum mass flow rate at the right side of the track where the train runs in Case 3 is larger by 10 kg/s than that in Case 1 when t=62s. The mass flow rate at the right side of the other track shown in Fig.8(c) isalways smaller than 0 (exhaust flow). Because of the existence of the block, the airflow on the other track is less affected by the flow on the track where the train runs in Case 3.

Table 4 Mass flow of air through the three ventilation ducts and the left and right sides during 0s＜t＜75s in Case 1[13]

Tables 2, 3 and 4 give in detail the air mass flows through the ventilation ducts and the left and right sides in the three cases during 0s＜t＜75s . As for the ventilation performance through the ducts, the Case 2 with the design of combined ventilation ducts (see Fig.2(b) and Table 2) sees a higher mass flow for each duct as compared with Case 1 with the separated ventilations (see Fig.3(b) and Table 3). But, in consideration of the total mass flow through all the ducts, the total mass flow-in and flow-out in Case 2 is only 70% of that in the Case 3. Also, the Case 3 with the installation of partition blocks between the two tracks sees a higher ventilation performance by about 10% as compared with Case 1 without partition blocks as shown in Table 5.

Table 5 The total air amounts of the mass flow-in and flow-out through ducts in Case 2 and Case 3 as compared with those in Case 1[13]during 0s＜t＜75s

The amount (absolute value) of the air mass flow through the left and right sides can be used as an index indicating the strength of the pertinent air flow caused by the piston effect due to a train motion. The mass flow (the absolute value) through the left and right sides in Case 2 is larger by about 45% as compared with that in Case 3. This is in line with the fact that the ventilation performance through all ducts in Case 2 is inferior to that in Case 3. Also, the mass flow (absolute value) through the left and right sides in Case 3 is larger by 18% as compared to that in Case 1, which means that the existence of partition blocks in case 3 not only increases the air ventilation through the ducts but also enhances the piston effect due to a train motion.

The net mass flow-ins for the whole computational volume are 42.624 kg in Case 1[13], 0.048 kg in Case 2, and -0.924 kg in Case 3, respectively. This may show that the current analyses are very accurate. It is to be noted that in this numerical analysis the pressure is set to zero at the ground openings of all ducts and at the left and right sides of the tunnel. Also, as the initial condition, the air velocity in the tunnel is set to zero, which means that there is no persistent air flow caused by the piston effect due to a train motion. These initial and boundary conditions may be responsible for some numerical errors.

5. Conclusions

In this article, the ventilation performance of different ventilation strategies is examined to find a ventilation method with a high performance in a subway tunnel in association with a train motion in the natural ventilation environment. For different ventilation strategies with different structures of ventilation ducts and geometries of partitions, computational analyses are carried out for train-induced unsteady air flows using the dynamic layering method for the treatment of the moving boundary of a running train.

The amount of air flow through a combined duct is larger than that through a separated duct, but is smaller than through two separated ducts, resulting in an inferior ventilation performance as a whole. In the case with several partition blocks installed between the two tracks, the airflow in the other track is less affected by the flow on the track where the train runs, resulting in an increase in the air ventilation through the ducts and the enhancement of piston effect due to the train motion.

The study results are applicable to the design of ventilation ducts and possible partitions between two tracks for the improvement of air quality in subway tunnels. Also, this work can provide some insight for pollutant behavior and fire disaster in a subway environment.

[1]ROH J. S., RYOU H. S. and PARK W. H. et al. CFD simulation and assessment of life safety in a subway train fire[J]. Tunnelling and Underground Space Technology, 2003, 24(4): 447-453.

[2]FUKUYO K. Application of computational fluid dynamics and pedestrian-behavior simulations to the design of task-ambient air-conditioning systems of a subway station[J]. Energy, 2006, 31(5): 706-718.

[3]YUAN F. D., YOU S. J. CFD simulation and optimization of the ventilation for subway side-platform[J]. Tunnelling and Underground Space Technology, 2007, 22(4): 474-482.

[4]KE M. T., CHENG T. C. and WANG W. P. Numerical simulation for optimizing the design of subway environmental control system[J]. Building and Environment, 2002, 37(11): 1139-1152.

[5]LI Li, DU Guang-sheng and LIU Zhen-gang et al. The transient aerodynamic characteristics around vans running into a road tunnel[J]. Journal of Hydrodynamics, 2010, 22(2): 283-288.

[6]KIM J. Y., KIM K. Y. Experimental and numerical analyses of train-induced unsteady tunnel flow in subway[J]. Tunnelling and Underground Space Technology, 2007, 22(2): 166-172.

[7]KIM J. Y., KIM K. Y. Effects of vent shaft location on the ventilation performance in a subway tunnel[J]. Journal of Wind Engineering and Industrial Aerodynamics, 2009, 97(5-6): 174-179.

[8]JIA L., HUANG P. and YANG L. X. Numerical simulation of flow characteristics in a subway station[J]. Heat Transfer-Asian Research, 2009, 38(5): 263-274.

[9]ZHAO Xiao-li, SUN Zhen-xu. A new method for numerical simulation of two trains passing by each other at the same speed[J]. Journal of Hydrodynamics, 2010, 22(5): 697-702.

[10]HUANG Yuan-dong, GAO Wei and KIM Chang Nyung. A numerical study of the train-induced unsteady air flow in a subway tunnel with natural ventilation ducts using the dynamic layering method[J]. Journal of Hydrodynamics, 2010, 22(2): 164-172.

[11]UDAYKUMAR H. S., MITTAL R. and RAMPUNGGOON P. et al. A sharp interface Cartesian grid method for simulating flows with complex moving boundaries[J]. Journal of Computational Physics, 2001, 174: 345-380.

[12]LIN C. J., CHAUH Y. K. and LIU C. W. A study on underground tunnel ventilation for piston effects influenced by draught relief shaft in subway system[J]. Applied Thermal Engineering, 2008, 28(5-6): 372-379.

[13]HUANG Y., HONG T. and KIM C. N. A numerical simulation of train-induced unsteady airflow in a tunnel of Seoul subway[J]. Journal of Mechanical Science and Technology, 2012, 26(3): 785-792.

[14]LAUNDER B. E., SPALDING D. B. The numerical computation of turbulence flows[J]. Computer Methods in Applied Mechanics and Engineering, 1974, 3(2): 269-289.

February 12, 2011, Revised March 28, 2011)

* Biography: HUANG Yuan-dong (1964-), Male, Ph. D., Professor

KIM Chang Nyung,

E-mail: cnkim@khu.ac.kr