You are on page 1of 16

Process Safety and Environmental Protection 147 (2021) 146–161

Contents lists available at ScienceDirect

Process Safety and Environmental Protection


journal homepage: www.elsevier.com/locate/psep

Numerical simulation of fire smoke control methods in subway


stations and collaborative control system for emergency rescue
Kai Wang a,b,c,∗ , Weiyao Cai b , Yuchen Zhang b , Haiqing Hao b , Ziting Wang b
a
State Key Laboratory of Coal Resources and Safe Mining, China
b
School of Safety Engineering, China
c
Key Laboratory of Coal Methane and Fire Control, Ministry of Education, China University of Mining & Technology, Xuzhou, China

a r t i c l e i n f o a b s t r a c t

Article history: As subway fires can easily cause numerous casualties, this paper proposes the idea of active disaster
Received 16 July 2020 relief, which involves ventilation, exhaust, smoke blocking and exhaust, and other facilities coordinating
Received in revised form 29 August 2020 centralised control to assist passenger evacuation and firefighting. The ventilation and smoke exhaust
Accepted 14 September 2020
method used in multilayer crossing complex subways is analysed, and a typical interchange transfer
Available online 18 September 2020
station is selected to build a numerical model. The combined control of fire smoke is analysed based
on six ventilation modes, and fire dynamics simulator software is used to perform several numerical
Keywords:
simulations. According to the simulation results, ventilation and smoke control is used in the cases of
Subway fire
Smoke isolation
fire on the second and third floor in the underground, the airflow speed of the stairway reached 2 m/s,
Mechanical smoke extraction which can completely restrain the smoke on the fire floor. According to the characteristics of different fire
Information fusion sources and smoke control scenarios, a remote monitoring model of multielement information fusion,
Collaborative centralised control such as ventilation path, fan characteristics, smoke exhaust channel, and smoke blocking facilities, is
established, and a multielement disaster information-fusion-based collaborative centralised emergency
rescue platform for subway fire disaster relief is developed. Real-time disaster analysis is performed,
and intelligent control of airflow and smoke exhaust in the complex form of subway-controlled smoke
turbulence field is developed to facilitate safe escape and efficient firefighting and rescue of people during
emergencies.
© 2020 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction experiences in subway operation and management is insufficient,


and the construction of the disaster emergency rescue system is
Outstanding features and advantages such as large passen- imperfect, resulting in the risk of major casualties in the event
ger volume, high speed, punctuality, convenience, safety and of a fire in the subway system (Zhong et al., 2016; Wang et al.,
environmental protection, have made the subway the preferred 2017a). Statistics show that about 85 % of subway casualties are
transportation mode for major cities to solve traffic congestion attributed to the hot and toxic smoke during a subway fire (Guo
(Cong et al., 2017; Wang et al., 2016a; Li et al., 2016; Weng et al., et al., 2018). Among them, the cross transfer station fire has its
2016). Several mega cities around the world, including Paris, Tokyo, particularity due to its structure, function, and the characteristics
New York, and London, have built an integrated and interconnected of people flow. Compared with a single station, there is a greater
subway network (Barbato et al., 2014; Se et al., 2012; Zhou et al., risk of fire in the transfer station, which is mainly manifested in
2017; Costantino et al., 2014). At present, 32 cities in China have the transfer station has complex functions and many factors that
subway networks that are either completed or under construction. cause a fire (Yuan and Smith, 2015; Ura et al., 2014). The under-
The Beijing Subway has become one of the busiest subways in the ground space structure of the transfer station is complex, and the
world (Ji et al., 2017; Yuan et al., 2018). In 2020, the total mileage of smoke spread in the case of fire is affected by many factors (Gao
the subway will reach 6100 km in China. Nevertheless, the general et al., 2016; He et al., 2017). Typical fire accidents in subways in the
recent years is shown in Table 1. All in all, the frequent occurrence
of fire accidents in subways is an important factor restricting the
development of the subway transportation. Statistical data show
∗ Corresponding author at: School of Safety Engineering, China University of Min- that an efficient emergency response system can reduce accident
ing & Technology, Xuzhou, 22116, China. losses to 6 % (Wang et al., 2016b). Subway ventilation and smoke
E-mail address: wangkai850321@163.com (K. Wang).

https://doi.org/10.1016/j.psep.2020.09.033
0957-5820/© 2020 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 147

Table 1
Typical fire accidents in subways in the recent years.

Time Site Occurrence reason Deaths Injuries

2000.11 Austria Train fire in the subway tunnel 155 18


2001.8 Sao Paulo, Brazil Fire in the subway station 1 27
2003.2 Taegu Fire in the subway station 487 146
2004.2 Moscow, Russia Explosion caused fire in the subway station 50 100
2011.4. Minsk, Belarus Explosion in the subway station 15 200
2012.11 Guangzhou, China Power failure fire 0 7
2013.6 Moscow, Russia Cable fire in the subway station 0 66
2014.7 Pusan, Korea Train equipment fire 0 5
2016.1 Tokyo, Japan Fire in the subway station 0 2
2017.2 Hong Kong, China, Fire in the subway station 1 18
2018.1 Guangzhou, China Cable fire in the tunnel 3 0
2020.3 New York, USA Train fire in the subway station 1 16

exhaust are the complex engineering systems, the process safety are emerging, posing new challenges to ventilation safety and dis-
and failure must be monitoring on the real time. In order to predict aster recovery. With the increasing emergence of complex forms
the failure, a dynamic Bayesian network offers much flexibility in of subway tunnels, such as ultra deep buried, multi-line crossing,
representing different failure scenarios and the interdependence large long interval, and dense platform, new challenges are put for-
of failure causes (Tanjin Amin et al., 2018). The applicability of ward to ventilation safety and disaster rescue. The fire development
deterministic modeling approach to model domino effect evolu- and smoke transportation behavior of the complex controlled flow
tion path, evolution times and evolution probabilities for different field in subway system, and the mechanism of multi-field coupling
primary fire accident scenarios, the approach is effective to model dynamic control of airflow and smoke in different scenarios need
the dynamic evolution of accident propagation (Ding et al., 2020a, to be further studied. The coordinated and optimized configuration
2020b, 2019). In order to prevent fires and mitigate their impact, the of multi-element air control and smoke exhaust facilities is the key
ranges of safety measures of fires are also studied by analyzing 14 to achieve efficient air control and smoke exhaust in subway fire.
credible scenarios (Yang et al., 2018). For the security reasons, the Intelligent control of complex forms of subway ventilation facilities
successful intrusion probabilities and security potentials of barriers and real-time monitoring and early warning of fire parameters is a
using a Bayesian network considering the dependency of barriers powerful guarantee for safe transportation.
and interaction of different intrusion scenarios (Song et al., 2018).
Inherently safer design is an effective solution to prevent accidents 2. Modelling and numerical simulation analysis of subway
because it postulates that the best way to reduce risk is to avoid the station fire
hazard rather than control the hazard (Abidin et al., 2018). Thus,
it is imperative to study the real-time evaluation of the subway The internal structure of a subway transfer station is complex,
disaster, including automatic airflow control and smoke exhaust and a large number of electrical devices support the safe opera-
technology. A collaborative centralised control system should be tion of the subway system. Once a fire occurs, it is difficult for
established, which combined mechanical ventilation and smoke passengers to escape. Therefore, the physical structure, ventilation,
isolation. Based on multielement disaster information fusion, a col- and smoke exhaust facilities of the subway station are analysed; a
laborative emergency rescue platform of subway fire disaster relief numerical model similar to the actual situation is developed, and
should be constructed, and a software and hardware system for the ventilation and smoke exhaust control conditions of the sub-
smoke flow control was developed, and a new emergency rescue way are determined together with the fire scene for the simulation
system was constructed. and analysis of subway fire.
For the fire smoke control in the subway station, the addition
of a centralized smoke exhaust will exhaust part of the smoke and
2.1. Subway station fire modelling and scene analysis
heat and induce longitudinal ventilation velocity, thus affecting the
critical velocity. The performance of the train fire screen door and
1) Analysis of subway station structure
ventilation system was evaluated by conducting an evacuation sim-
ulation and verified that the platform screen (PSD) and effective
ventilation equipment in the subway fire can improve the evac- Herein, the typical “cross” transfer station fire is presented as an
uation efficiency (Krol et al., 2017; Tanaka et al., 2015). Several example to study the smoke movement law and control method.
mechanical ventilation forms under tunnel fire conditions were The control of subway disaster smoke flow mainly depends on
simulated and compared by Computational Fluid Dynamics (CFD) ventilation, the smoke exhaust system, ceiling screen, and other
(Gannouni and Ben, 2015). The temperature, visibility, and CO con- facilities. The typical structure of a subway ventilation and smoke
centration distribution at the height of the human eye (1.6 m) in exhaust system is shown in Fig. 1. The spatial structure system of
the subway station were studied under different smoke exhaust the Beijing Huang Zhuang metro station is shown in Fig. 2.
modes, and determined the smoke exhaust mode that meets evac- Fig. 1 shows the composition of the subway ventilation and
uation requirements (Muduli et al., 2018). The effects of different smoke exhaust system in detail. The specific principles are
ventilation methods were studied for the safe evacuation of per- described as follows. Under the power of the fresh air blower, the
sonnel (Mei et al., 2017). Thus, the transport characteristics of air flow enters the mixed air chamber from the air inlet pavilion
the complex form of subway fire smoke flow under the action of through the muffler, and the combined air-conditioning cabinet
multiple facilities should be analysed, such as ventilation, smoke provides fresh air flow to the platform, station hall and other
exhaust, smoke blocking, and isolation. Numerical model of a typ- spaces through the air supply pipe network. Under the power of the
ical cross-interchange station, and different fire scene of smoke return/exhaust fan, the exhaust subway station will be discharged
control modes should be constructed, from the ground by the exhaust pavilion through the platform and
Complex forms of subway tunnels such as ultra-deep burial, the return exhaust pipe network of the station hall, to realize air
multivariate line crossing, large long interval, and dense platform replacement in underground space. In the event of fire, in order to
provide support for the evacuation of personnel,the smoke flow is
148 K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161

Fig. 1. Structural schematics of subway ventilation and smoke exhaust system.

Fig. 2. Spatial structure schematics of actual operation of Xin Zhuang metro station.

isolated and the smoke exhaust fan is turned on to quickly discharge lines 4 and line 10 cross, a simplified model was established: the
the smoke flow from the ground through the smoke blocking arm geometric dimension of the station hall layer was set as 128 m
and other facilities of the fire compartment. Fig. 2 shows the spatial long, 18 m wide, and 5 m high; the two platform layers were 128 m
structure of the Beijing Subway Huang Zhuang interchange station, long, 16 m wide, and 5 m high. The numerical model of the specific
which is a typical c̈rossẗransfer station,the setting of entrances and subway station are shown in Fig. 3.
exits and the relationship between layers guide the construction of The ventilation and smoke exhaust system facilities were con-
physical models for fire smoke simulation in subway stations. structed as follows: the ventilation system of the subway station
was on the first, second, and third floors in the underground; every
2) Metro station modelling and parameter setting floor comprised two air supply pipes, and two return/exhaust pipes
were arranged along the length direction. In the normal ventilation
mode, the airflow speed of the air inlet and air outlet was 1 m/s; in
To simulate the smoke flow control of subway fire more realisti-
the disaster state, the air outlet was connected to a smoke exhaust
cally and to enable a rapid rescue after a disaster, the actual subway
fan, the exhaust speed was 2 m/s, and the air supply speed was 1
structure was simplified, where some equipment and structures
m/s. The number and air volume of the inlet and exhaust outlets are
that were insignificant to the simulation results were disregarded;
shown in Table 2. Together with the numerical model of the metro
subsequently, the Pyrosim numerical model of the “cross” transfer
station, the location features of the air inlet and exhaust outlets are
station was established.
shown in Fig. 4.
This model comprised three layers: the second and third floors in
The ceiling screen in the public area of the metro station was set
the underground were platform layers, the first floor in the under-
at the top of the station hall and platform and around the stairway.
ground was a station hall layer, and multiple sets of stairs were
Each station hall floor and platform floor comprised two smoke
set between the station hall layer and platform layer. According to
control zones, each of which measured 600 m2 in area. A ceiling
the design of an island station, the two sides of the platform layer
screen was set at the edge of the smoke control zone to block and
were set as the track area, whereas the track area and platform area
gather smoke flow. The sagging height of the ceiling screen was 2 m
were separated by symmetrical screen doors on both sides, and the
such that the smoke storage height of the ceiling was 2 m, which can
screen doors were closed to isolate smoke in case of disaster to
effectively prevent smoke from spreading to the adjacent smoke
improve the smoke removal efficiency of the platform. Based on
control zones.
the investigation of the Huang Zhuang station where Beijing Metro
K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 149

Fig. 3. Schematics of station platform.

In terms of combustibles in a conventional fire, the combustibles rate of 1.2 MW in experiments with handcart piled with multiple
in the platform included mainly those in luggage and garbage cans. pieces of luggage (Ingason, 2008). In this paper, the heat release rate
The statistical research results of Hansell and Morgan show that the of the steady-state combustion of the fire source is set as 1.5 MW,
fire intensity in densely populated places is about 2 MW. The UK’s considering that the subway station has a high density of personnel
Fire Research Institute (FRS) measured the maximum heat release and many pieces of luggage are continuously ignited.
150 K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161

Table 2
Smoke exhaust and air supply parameters.

Location Number of Normal exhaust Disaster exhaust Number of air Supply air rate
smoke outlets air rate (m3 /h) air rate (m3 /h) inlets (m3 /h)

First floor in underground 26 93600 187200 48 172800


Second floor in underground 50 180000 360000 50 180000
Third floor in underground 44 158400 316800 44 158400

Table 3
Four kinds of grid scheme.

Scheme Grid size/m Number of Grids Run time/s

1 0.10 5000000 300h50min


2 0.15 2603670 190h31 min
3 0.20 1250000 80h18min
4 0.25 860000 40h4min

Q is the heat release rate, kW; a is the ambient air density,


kg/m3 ; CP is the specific heat of air, J/kg ◦ C TA is the ambient tem-
perature, K; g is the acceleration of gravity, m/s2 .
The FDS technical reference guide summarises the parameter
settings of 36 types of numerical simulations of different building
structures. Most of the values of D*/␦x are between 3 and 16. Some
scholars believe that when the value of D*/␦x is between 4 and 16,
the numerical model can solve the turbulence problem in the fire
Fig. 4. Schematics of air supply and exhaust outlets.
smoke well (Wang et al., 2016c). Currently, a value between 10 and
20 is still used in some studies. In this study, MATLAB was used
Grid size is a parameter that must be set in a numerical simula- to calculate the relationship among various D*/␦x values, different
tion. Because of the significant amount of calculation in small-scale fire heat release rates, and mesh resolutions, as shown in Fig. 5.
eddy current simulation, the basic turbulence equation must be Fig. 5 shows that the grid is not sufficiently dense, and the accu-
filtered in the large eddy simulation method through Favre to deter- racy is relatively insufficient when selecting a 3X scale. Selecting a
mine the large and small eddy structures, which must be nested 20X scale results in an extremely dense grid, which would be com-
in the subgrid (Smagorinsky) model. The grid size setting must putationally time consuming. Therefore a 4X to 10X scale is more
capture the large eddy structure in the flow field. reasonable. According to the simulated fire source power of 1.5
In the fire dynamics simulator (FDS) user guide, Mc Grattan MW, ␦x was calculated to be between 0.11 and 0.27. Choose four
stated that the ratio of the characteristic diameter of the fire source grid sizes within this range, and verify the grid independence of
to the nominal size of a mesh cell D*/␦x represents the precision of the simplified model calculation results of typical simulation con-
the fire source grid, in which a greater value implies a higher pre- ditions. The simulation time is set to 360 s. The specific scheme
cision (McGrattan and Forney, 2018). D* is expressed as follows: is shown in the Table 3. Combined with the verification of grid
independence, the grid size selected in this simulation was 0.15
2
m.
Q 5 Fig. 5b shows the grid setting of the local encryption area around
D∗ = ( √ )
˛ cp TA g the fire source. For the grid setting of other areas, refer to the results
of the grid accuracy verification which is calculated, the grid size

Fig. 5. Effects of D*/␦x and grids selection demonstration.


K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 151

Fig. 6. Schematics of parameter measuring point setting.

Table 4
Ventilation and control modes of different cases.

Case Fire location velocity of single velocity of single Ventilation modes Control mode
air inlet smoke outlet

1 1 m/s 1 m/s Normal ventilation The air supply and exhaust outlets of
Third floor in the
each floor are opened
underground platform
2 1 m/s 2 m/s Ventilation control mode 1 Open the exhaust on the third floor in
the underground; open supply outlet
on the first floor
3 1 m/s 2 m/s Ventilation control mode 2 Open the exhaust on the third floor in
the underground; open supply outlet
on first and second floors
4 Second floor in the 1 m/s 1 m/s Normal ventilation The air supply and exhaust outlets of
underground each floor are opened
5 platform 1 m/s 2 m/s Ventilation control mode 3 Open the exhaust on the second floor
in the underground; open supply
outlet on the first floor
6 1 m/s 2 m/s Ventilation control mode 4 Open the exhaust on the third floor in
the underground; open supply outlet
on the first and third floors in the
underground

of the area away from the fire source is generally twice that of the case conditions were established for numerical simulation, and the
fire source area (Weng et al., 2014; Gao et al., 2015). movement rules before and after the fire smoke flow control were
In the FDS simulation, many monitoring points were established observed comprehensively. In all cases, the ceiling screens were
to obtain smoke parameters at specific locations to analyse the tem- in the drooping state, and the entrance of the station hall was in
poral and spatial evolution of fire smoke flow. According to the an open-ventilation state. The specific parameters of the six fire
characteristics of subway fire and the people’s needs in an escape control air and smoke exhaust scenarios are shown in Table 4.
environment, a thermocouple was placed every 1 m at the top of A brief introduction to the six cases is presented as follows: All
the fire site, whereas a CO sensor, visibility sensor, and tempera- air supply and exhaust outlets in cases 1 and 4 were opened to
ture sensor were placed at the height of the human eye (1.6 m). maintain normal ventilation, and no pressure difference existed
Furthermore, sensors for measuring the airflow speed, smoke, CO between the layers. In case 2, the smoke exhaust fan of the third
concentration, and visibility were placed at all stairways. Addition- floor in the underground was turned on to generate a negative pres-
ally, those sensors were placed at the junctions of the second and sure; the air inlet of the first floor in the underground was opened
third floors in the underground as well as at the height of human to generate a positive pressure, which produced an airflow from the
eyes in the centre of the smoke prevention zone on each floor. The first floor in the underground to suppress smoke. In case 3, the air
arrangement of the sensor measuring points on both sides of the supply outlet on the second floor in the underground was opened,
ignition point on the third floor in the underground is shown in as in case 2, to increase air supply. In case 5, the smoke exhaust
Fig. 6. fan of the second floor in the underground was turned on to gen-
erate a negative pressure, and the air inlet of the first floor in the
3) Scene analysis and settings underground was opened to a generate positive pressure, which
produced an airflow from the first floor in the underground to the
second floor in the underground to suppress the smoke. Based on
The location of the subway fire was random selected; the ven- case 5, the air supply outlet of the third floor in the underground
tilation and smoke exhaust methods were various, and the escape was opened in case 6 to increase air supply.
path was complex. Therefore, it was difficult to evacuate and extin-
guish the fire after the disaster. To master the control method of
subway disaster smoke flow and improve the efficiency of venti- 2.2. Numerical simulation results of fire scene in MTR Station
lation and evacuation, the flow organization mode of fire smoke
and fresh air in the public area of the platform, the pressure distri- The application range of the fire dynamics simulation software
bution characteristics of the station hall and each platform layer, FDS was extended from ultrahigh buildings, enterprise workshops,
and the effects of the operation states of the air supply unit and air etc. to roadways and subways, and the abstract fire smoke evolu-
return unit on the smoke movement were investigated by setting tion into a concrete image was investigated successfully. Combined
different ventilation and smoke exhaust facilities in the numeri- with the ventilation and smoke exhaust system in a subway station,
cal simulation. Based on the analysis of the subway fire scene, six the distribution law of smoke flow, temperature, and visibility of
152 K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161

Fig. 7. Snapshots of smoke spread in case 1.

different fire sources in the normal air supply and smoke exhaust
fan system was investigated in this study to establish an evacuation
route and plan for fire rescue.
High concentration smoke produced by fire not only affects the
breathing of passengers, but also reduces the visibility of the evac-
uation path, which significantly reduces the evacuation efficiency.
According to the metro design code, the evacuation time shall not
exceed 6 min. The FDS numerical simulation time was set to 400 s,
and the simulation was performed according to the scene settings.
The distribution rule and evolution path of the high concentration
smoke produced by the fire in the subway structure was analysed
to establish a smoke flow control method and personnel evacuation
path selection. The numerical simulation results of the fire smoke
migration in case 1 are shown in Fig. 7. The average airflow speed
of stairs 1–3 and stairs 2–3 in case 1 is shown in Fig. 8. The smoke
diffusion condition of cases 2 and 3 at 360 s is shown in Fig. 9. The
average airflow speed of each stair in cases 1, 2, and 3 is shown in
Fig. 8. Average airflow speed of stairs 1–3 and 2–3 in case 1.
Fig. 10. The ceiling temperature in the horizontal direction of the
fire source in cases 1, 2, and 3 is shown in Fig. 11.
As shown in Fig. 7, when the case 1 platform layer begins to catch outward, and the temperature decreased after the smoke spread
fire, the fire source continuously produced smoke as the fire pro- when entraining the air. Hence, the resulting thermal buoyancy
ceeded. At the beginning, the smoke exhaust rate of the exhaust decreased, and the buoyancy of the smoke became smaller than that
fan was lower than the rate of the smoke produced by the fire of gravity, which resulted in settlement. When the smoke sank over
source. Consequently, the smoke accumulated slowly and spread the ceiling screen, it spread upward. At this time, no strong airflow
K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 153

Fig. 9. Smoke diffusion condition of Cases 2 and 3 at 360 s.

Fig. 10. Average wind speed of each stair in case 1, 2, and 3.


154 K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161

Fig. 11. Ceiling temperature in horizontal direction of fire source of cases 1, 2, and 3.

flowing down the staircase was available to blow back the smoke; subsequently, the airflow speed fluctuated around a certain value.
hence, the smoke spread to the upper layer because of buoyancy, In cases 2 and 3, owing to the combined effect of the exhaust fan,
endangering the safety of the upper personnel. In case 1, the smoke ceiling screen, and airflow control smoke at the stairway, the smoke
had spread to the second and third floors in the underground of the was firmly controlled at the fire source. To achieve this balance, the
staircase at 200 s, and some of the smoke had spread to the second airflow and ceiling screen operated synergistically to gather the
floor in the underground. The connecting aisle from the second floor smoke and improve the smoke exhaust efficiency. At this time, the
to the first floor in the underground had already been filled with smoke exhaust rate of the smoke exhaust was equivalent to the rate
smoke. At 400 s, the smoke had spread through stair 1–3 into the of smoke generated by the fire source; therefore, the smoke could
first floor in the underground of the station hall, a large part of the not spread to other locations. After the air supply was turned on
station hall side of the space was filled with smoke, the second floor in case 2, the airflow on the first floor in the underground station
in the underground was partially filled with smoke, and the smoke hall passed through the second floor stairs, and control the smoke
have the trend to fill the first and second floors in the underground. on the third floor in the underground. The average airflow speed of
As shown in Fig. 7, the smoke exhaust efficiency of case 1 was the six measuring points at the stairway exit oscillated at 1.4 m/s.
insufficient. Even though a ceiling screen was installed to collect And the wind speed of the two stairs connecting the first floor to
smoke to improve the smoke exhaust efficiency, the smoke exhaust the third floor (No.1 stair 1–3 and No. 2 stair 1–3) is about 0.9 m/s,
rate was still lower than the rate of smoke produced by the combus- which also can prevent the smoke spreading to the other floor. In
tion of the fire source, and the smoke was increasing. The airflow case 3, the air supply outlets of the second floor platform to the first
speed at the stairway could not stop the spread of smoke, and the floor in the underground were opened. The opening of the second
smoke moved to the upper station hall. floor vent in the underground was equivalent to increasing the air
As show in Fig. 8, because the air outlets and air supply openings flow from the first floor to the third floor in the underground. The
of each floor were opened simultaneously in the normal ventila- monitoring of the measurement points at the stairway entrance
tion mode, the negative pressure of the air outlet and the positive indicated that the airflow speed fluctuated at approximately 2 m/s.
pressure of the air outlet cancelled each other out, and the pres- And the airflow speed at the two stairs connecting the first floor to
sures of the various floors did not differ significantly. The airflow the third floor (No. 1 stair 1–3 and No. 2 stair 1–3) is around 1 m/s,
speed due to the pressure difference was low as well. The results which also can limit smoke on fire floor.
of the measurement points indicate that the airflow speed on the The higher airflow speed of case 3 enabled a better smoke con-
side of the fire was less than 0.6 m/s, which could not prevent the trol; therefore, smoke was concentrated in the smoke prevention
smoke from spreading to the upper layer; therefore, case 1 venti- zone, and the concentration of smoke increased, thereby increasing
lation mode cannot be used when a fire occurs on the third floor in the working efficiency of the smoke exhaust port. Furthermore, the
the underground. amount of smoke pulled out was larger and hence the effect of case
According to the simulation results, Fig. 9 shows that both cases 3 was better than that of case 2.
2 and 3 can control the smoke from spreading upward, and that Fig. 11(a) shows the thermocouple measurement results at 0.2
the smoke control effect is significantly better than that of case 1. m from the ceiling of the third floor in the underground, whereas
The smoke in both cases was always on the fire floor, and there Fig. 11(b) shows those after the removal of several points around
was no sign of smoke spreading on the two stairs of the fire floor. the fire source and temperature difference amplification. Owing to
The smoke control effect of case 3 was better than that of case 2. the airflow effect of the staircase between the second and third
A small amount of smoke spread out of the smoke-proof zone on floors in the underground, the flame plume was not perpendicular
the side of the fire. In this case, the subway had sufficient smoke- to the top. Because of the airflow effect of the stairwell between the
exhausting capacity, and the spreading smoke was pulled into the second and third floors in the underground, the flame plume was
smoke exhaust at the stairs; it could not accumulate or spread to not perpendicular to the upper side but biased to one side, resulting
the upper floors. in the temperature above the fire source not being the highest. The
As shown in Fig. 10, the pressure difference formed by the fans in temperature in case 1 on the left ceiling near the fire source was
the two fire ventilation modes caused the airflow speed to increase;
K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 155

Fig. 12. CO distribution of cases 1, 2, and 3 at eye height (1.6 m) at 100 and 360 s.

the highest because smoke could not be effectively removed and resulting in a lower CO concentration in the left space. The blue area
the temperature was high. on the left was the downward airflow of stairs 1–3, which caused
The temperature at the high-temperature zone was disregarded, CO to be blown away and the concentration to be lower than the
and the measurement points around the high-temperature zone surrounding concentration.
were selected for comparison. The temperatures of cases 2 and 3 The temperature distribution shown in Fig. 13 shows that the
were similar, whereas that of case 1 on the side of the fire source downward airflow of stair 1–3 blew to the wall such that the air
was approximately 5 ◦ C higher than those of the other two cases, flowed in the opposite direction to the wall. Because smoke control
which indicates that the two fire ventilation modes exhausted more and ventilation measures were implemented, the high tempera-
heat relative to case 1 normal ventilation mode and resulted in a ture area was controlled in the smoke control zone, indicating that
significant temperature decrease. the mechanical smoke exhaust system of cases 2 and 3 effectively
To assess the safe escape path for subway passengers and to eliminated and suppressed the smoke and reduced the smoke tem-
analyse the migration and distribution of hot smoke airflow in dif- perature. The temperature of the large area around the fire point
ferent cases, the CO concentration distribution was analysed, as was below 50◦ , which was conducive to evacuation. However, the
shown in Fig. 12; moreover, the temperature and visibility distri- high-temperature red area of case 3 was smaller than that of case
butions are shown in Figs. 13 and 14, respectively. 2. Only one side exhibited a higher temperature; therefore, case 3
Fig. 12 shows that under the action of many exhaust fans, a large was better than case 2 for temperature control.
portion of the CO produced by the fire was exhausted by the fan, As shown in Fig. 14, the visibility distribution rules of fire ven-
and its high-concentration area was near the fire source and the fire tilation modes for cases 2 and 3 were similar. The smoke was
layer, which was close to the corner position. Under the effects of controlled in the smoke-control zone on the left side; a small
the first and second floors in the underground downward airflow, amount of smoke overflow was blown back by the airflow under the
CO gathered to the lower left. Because there were fewer vents in stairway entrance; and no smoke entered the second floor in the
the corners, the smoke could not be effectively exhausted. underground. The fire ventilation mode of cases 2 and 3 controlled
Although the distribution rules of cases 2 and 3 were similar, the the smoke better than that of case 1. The red area on the left side
smoke was concentrated at the leftmost end. However, the down- was stair 1–3, where the downward airflow blew the smoke away,
draft of case 3 rendered the exhaust efficiency higher than that resulting in a higher visibility compared with the surrounding area.
of case 2. Furthermore, CO was removed with the smoke, thereby
156 K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161

Fig. 13. Ceiling temperature of cases 1, 2, and 3 at 100 and 360 s.

Based on the results, case 1 ventilation and smoke exhaust mode As shown in Fig. 15, case 4 was in the normal ventilation state,
is not applicable to fires on the third floor subway platform in and the smoke spread was not controlled. At 360 s, the smoke
the underground. The maximum downward airflow speed of the spread from the second floor to the first floor. By contrast, the smoke
stairway was only 0.6 m/s, which could not effectively control and was controlled better in cases 5 and 6. Within 360 s, the smoke was
eliminate smoke. In the simulation, part of the smoke spread to the controlled in the smoke control zone on the fire side and did not
second floor in the underground. In cases 2 and 3, the smoke was spread to other floors.
concentrated in the fire area, and the CO, temperature, visibility, As shown in the pressure and airflow speed distribution dia-
and other indicators were significantly better compared with case grams in Figs. 16 and 17, because the air inlet and exhaust vents
1. The downward airflow speed of case 2 stairway was less than that were all opened during case 4 fire, the positive and negative pres-
of case 3, and the ability to collect and remove smoke was slightly sures on the same floor cancelled each other out, and the pressure
less. The CO and temperature distribution diagram shows that case difference between the fire floor and other floors was small. This
2 was weaker than case 3. Therefore, case 3 should be considered situation is similar to case 1 and will not be repeated. In case 5, the
for controlling airflow and exhaust smoke for fires occurring on the exhaust port was opened on the second floor in the underground
third floor in the underground. to provide a negative pressure, whereas the air supply port was
The fire sources of cases 4, 5, and 6 were set on the side of the opened on the first floor in the underground to generate a positive
second floor in the underground platform. According to the prede- pressure. The pressure on the first floor in the underground was sig-
termined ventilation control and smoke exhaust mode, the smoke nificantly higher than those on the 2nd floor in the underground,
migration diagram 360 s after the occurrence of fire is shown in and the downward airflow at the stairway prevented the smoke
Fig. 15. The pressure distribution diagram of each layer is shown from spreading upward.
in Fig. 16; the airflow speed distribution characteristics of the key In case 6, the pressure was positive owing to the simultaneous
stairway entrance are shown in Figs. 17 and 18, whereas the tem- supply of air from the first and third floors in the underground and
perature distribution curves of the ceiling is shown in Fig. 19. The the negative pressure from the exhaust of the second floor in the
co-distribution characteristic curves at a 1.6 m height are shown in underground, which increased the airflow speed of the third floor
Fig. 20.
K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 157

Fig. 14. Visibility of cases 1, 2, and 3 at 100 and 360 s.

in the underground to the second floor (compared with case 5),


thereby restricting the smoke flow in the fireproof area.
As shown by the monitoring data presented in Fig. 18, the airflow
speed changed with time in three cases. The pressure difference
between the floors in different cases resulted in a large difference
in the airflow speed at the staircase. Among the three staircases
connecting the second and first floors in the underground in case
5, the difference between the two layers was the largest, and the
airflow speed of the three staircases was the highest, with average
values fluctuating at about 1.75, 1.5, and 1.5 m/s. In case 6, the air
supply port of the third floor in the underground was opened such
that the third floor in the underground was in a positive-pressure
state. The airflow speed at the stairway was smaller than that of
case 5, and the values were 1.2, 1.0, and 1.3 m/s. The airflow speed
limited the spread of the 1.5 WM fire smoke. In case 4, almost no
pressure difference was observed in any floor; therefore, the airflow Fig. 15. Smoke diffusion condition of cases 4, 5, and 6 at 360 s.
speed was the smallest, and the airflow speed change was unstable,
pulsating at approximately 0.25 m/s. ports was small, the amount of smoke exhausted was less than that
As show in Fig. 19, the spread of smoke in case 5 can be lim- of cases 4 and 6, and the heat extracted by the smoke exhaust was
ited when fire ventilation is enabled, but the ceiling temperature less. In the case 5 fire ventilation mode, a negative pressure had
is higher than those of the other two cases, especially at the right to be formed on the fire layer to limit the smoke. The air supply
side of the fire source; the closer to the fire source, and the gap is port was not opened as in case 4, and the lack of low-temperature
larger, the maximum reaches 40◦ C. In case 5, only the second floor fresh air flow reduced the smoke temperature. Because the smoke
smoke exhaust fan was turned on, the number of smoke exhaust was controlled in the smoke prevention zone of the fire area, the
158 K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161

Fig. 16. Pressure distribution of cases 4, 5, and 6 at 360 s.

Fig. 19. Ceiling temperature in horizontal direction of fire source of cases 4, 5, and
6.

Fig. 17. Velocity distribution of cases 4, 5, and 6 at 360 s.

heat was more concentrated compared with case 4, and the ceiling
temperature of case 5 was higher.
The three staircases on the second and first floors in the under-
ground were equipped with CO-measuring points. As shown from
the monitoring results in Fig. 20, a large amount of CO was detected
at stairs 1 and 2 in case 4, which is consistent with the smoke dis-
tribution map. The other two cases suppressed the flow of CO to
the upper layer in 360 s, and the CO concentration was low and Fig. 20. Volume fraction of CO in stairways of cases 4, 5, and 6.
negligible. Although the CO concentration differences in all cases
were considerable, they were all within the safe range, and were
thus not detrimental to the human body. to form a negative pressure; furthermore, the air inlet on the first
When a fire occurred, the normal ventilation of cases 4 and 1 was floor in the underground was opened to provide a positive pressure,
similar, and almost no pressure difference was observed between which generated a high speed of airflow at the stairway entrance,
the floors, which caused the airflow speed of the stairs on the first thereby restraining the smoke from spreading to the upper layer.
and second floors in the underground to be low; hence, the smoke As shown from the monitoring results of the roof temperature mea-
could not be prevented from spreading to the upper layer. In cases suring point at the fire source location, the roof temperature of case
5 and 6, the smoke was effectively controlled because the air inlet 5 was extremely high relative to that of case 6; as such, the tem-
was closed on the fire floor, whereas the smoke exhaust was opened perature cannot be well suppressed. Therefore, when a fire occurs

Fig. 18. Average airflow speed of each stair in cases 4, 5, and 6.


K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 159

Fig. 21. Structural block diagram of the subway fire collaborative control system.

on the second floor in the underground, case 6 provides the best and smoke flow also need to be deeply integrated. Based on the
ventilation mode. ES-DSS autonomous dynamic emergency rescue decision-making
model of complex forms of fires, an autonomous dynamic rescue
decision-making system of complex forms of subway fires is devel-
3. Research and design of fire smoke control system in
oped, using distributed measurement and control point layout and
subway station
collaborative centralized control methods to achieve intelligent
regulation of catastrophic air and smoke flow. The system consists
The subway fire smoke spreads fast, has strong toxicity, and it
of distributed monitoring substation, industrial Ethernet commu-
is difficult to evacuate when the fire occurs which is difficult to
nication module, environmental monitoring module, automatic fire
rescue, which poses a great threat to the life safety of passengers.
extinguishing spray module, remote screen door control module,
Therefore, the article study on the smoke flow control mecha-
remote smoke ceiling screen control module, supply fan and gate
nism of different fire locations in the complex subway network
valve control module, smoke exhaust fan, and It consists of gate
structure, and a collaborative centralized control system combin-
valve control module, fire door status monitoring module, ES-DSS
ing mechanical ventilation and smoke isolation is established. In
independent dynamic decision-making module, remote disaster
order to solve a real decision-making problem in a much more
relief command center and so on. The subway fire emergency res-
reliable and appropriate way, Yazdi introduces a novel approach
cue cooperative centralized control system, as a key equipment for
that integrates optimization features, DEMATEL (decision-making
controlling airflow and smoke in a catastrophic condition, must
trial and evaluation laboratory), and Bayesian structural network
start and operate stably and reliably and realize the control of the
(Yazdi et al., 2020). Analysis of the characteristics of subway fire
remote command center. The block diagram of the structure of the
scenes and ventilation and smoke exhaust devices to constructs
complex centralized control system for subway fire air and smoke
the expert system - decision support system (ES-DSS) autonomous
flow is shown in Fig. 21.
dynamic decision-making model for disaster relief, determines the
The related monitoring sub stations of each module of the
best scheme of airflow and smoke control, controls the polluted
complex fire air and smoke flow collaborative centralized control
area of smoke flow, and guarantees the fresh air supply to create
system collect data in real-time, transmit data to the monitoring
conditions for the rapid evacuation of passengers.
center, and the database of the monitoring center processes and
In a complex form of the transfer station, the air flow is com-
analyzes the ventilation status of the subway in real-time. Under
posed of multiple branches to form a network structure, and
normal conditions, real-time monitoring of ventilation and air con-
regulating a branch will affect the entire network structure. There-
ditioning of the subway system shall be carried out to ensure the
fore, the ventilation network structure of the multi-branch subway
comfort and safety of personnel. In case of abnormality, firstly,
station is studied. After the fire occurs, through the coordinated
obtain the fire information through the environmental monitor-
centralized control of smoke blocking, isolation, ventilation, and
ing module, and the monitoring center obtains the best ventilation
smoke exhaust facilities, the purpose of reaching the minimum
and smoke exhaust scheme by using the ES-DSS independent disas-
smoke pollution area and quickly exhausting the ground is reached.
ter relief decision-making system; secondly, the monitoring center
With the complexity of the subway station network structure
issues instructions to trigger the linkage control of the fire door
changing, new challenges have been raised for post-disaster emer-
and the closing of the ceiling screen, monitor the status of the air-
gency rescues such as smoke isolation methods, optimal emission
flow control facilities in real-time, control the fire smoke flow in a
paths, and optimal passenger evacuation path planning. Moreover,
small range, and start the sprinkler system, Quickly extinguish the
there are still no specific smoke control guidelines for such complex
fire source; finally, switch the relevant air valves and gate valves,
station.
open the air supply fan and open the smoke exhaust fan to realize
Subway integrated supervisory control system plays an impor-
the fresh air flow in and smoke exhaust. According to the distribu-
tant role in the daily operation of the subway and the early
tion of smoke and passage, the system adjusts the air control and
warning and rescue of disasters. The complex form of subway dis-
aster emergency decision-making and intelligent regulation of air
160 K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161

Fig. 22. Visual form of ES-DSS Decision Making System for Subway Disaster Rescue.

smoke exhaust scheme in real-time to create the best conditions can completely restrain the smoke spread. However, in case
for personnel evacuation and fire fighting and disaster relief. 1, the pressure difference of each floor was small under nor-
The integrated background computing software is built into mal ventilation, causing the airflow speed of the stairway to be
the cooperative control system of airflow and smoke flow in the lower than 0.6 m/s, which hindered the collection of exhaust
complex form of subway fire. All the monitoring parameters are smoke and prevented it from spreading to the upper layer. The
coordinated and analyzed to ensure the optimal ventilation under airflow speed of case 2 was 1.5 m/s, less than case 3, thereby
normal conditions. After the fire, according to the location and enabling smoke elimination. However, compared with other dis-
spread characteristics, the best evacuation route for passengers is aster indicators, case 2 exhibited lower CO removal capacity,
analyzed to determine the smoke flow path and the control method cooling capacity, and visibility in the smoke flow area compared
of fresh air flow under the premise of ensuring the safety of per- with case 3.
sonnel. Firstly, the ventilation network structure of the subway, the 2) Case 4 is similar to case 1 in terms of smoke flow characteristics.
working conditions of the air supply fan and the smoke exhaust The airflow speed of case 5 oscillated at 1.75, 1.5 and 1.5 m/s in
fan, and the ventilation path planning after controlling the air and three staircases. The airflow speed of case 6 oscillated at 1.2, 1.0,
smoke exhaust are embedded into the background database. After and 1.3 m/s in three staircases; both cases demonstrated good
the fire, the real-time air volume on the key air supply and smoke smoke exhaustion. In case 5, the number of opened air supply
exhaust path in the disaster relief process can be obtained through outlets was less than that in case 6, resulting in the poor cooling
the continuous iteration of airflow network calculation, which pro- effect of fresh airflow. The measuring point temperature of case
vides data support for the control of fire smoke flow. The complex 5 was significantly higher than that of case 6 (the maximum
form of subway fire specific flow collaborative control system by difference was 40 ◦ C), which affected the evacuation of people in
numerical simulation and field test, calculated the catastrophe in the fire layer. Therefore, case 3 and case 6 should be considered
the process of the ideal air volume, air exhaust path according to for controlling the air and exhaust smoke after a fire occurs on
the ideal air volume setting threshold, when the actual air volume the different floors in the underground.
and air volume ideal error more than 10 %, the system automati- 3) The characteristics of a subway fire scene, ventilation, and smoke
cally adjusts the frequency of the supply air exhaust fan, provide exhaust device were analysed. The expert system - decision
good conditions for evacuation. The remote visualization form of support system autonomous dynamic decision-making method
the complex form of the cooperative centralized control system for of disaster rescue was proposed, and the best scheme for air
subway fire and smoke flow is shown in Fig. 22. control and smoke exhaust was determined. An intelligent simu-
lation software was developed to study and evaluate the disaster
4. Conclusions situation in real time, dynamically convey the command of
smoke control and evacuation information, effectively control
1) Six of air control modes for fires in a typical cross-interchange the smoke pollution area, and ensure fresh air supply to facilitate
subway station were established. The numerical simulation rapid evacuation of passengers.
results showed that the airflow continuously flowed to the fire
floor, and the airflow speed of the stairway reached 2 m/s, which
K. Wang et al. / Process Safety and Environmental Protection 147 (2021) 146–161 161

Declaration of Competing Interest Li, Y., Fan, C., Ingason, H., et al., 2016. Effect of cross section and ventilation on heat
release rates in tunnel fires. Tunnelling Undergr. Space Technol. 51, 414–423.
McGrattan, K., Forney, G., 2018. Fire Dynamics Simulator (Version 6) User’s Guide.
The authors report no declarations of interest. Natl. Inst. Standards Technol.
Mei, F.Z., Tang, F., Ling, X., Yu, J.S., 2017. Evolution characteristics of fire smoke layer
Acknowledgments thickness in a mechanical ventilation tunnel with multiple point extraction.
Appl. Therm. Eng. 111, 248–256.
Muduli, L., Jana, P.K., Mishra, D.P., 2018. Wireless sensor network based fire moni-
This work was partly supported by the Fundamental Research toring in underground coal mines: a fuzzy logic approach. Process Saf. Environ.
Funds for the Central Universities (2019GF03), and the Priority Aca- Prot. 113, 435–447.
Se, C.M.K., Lee, E.W.M., Lai, A.C.K., 2012. Impact of location of jet fan on airflow
demic Program Development (PAPD) of Jiangsu Higher Education structure in tunnel fire. Tunnelling Undergr. Space Technol. 27, 30–40.
Institutions. Song, G., Khan, F., Yang, M., 2018. Security assessment of process facilities-Intrusion
modeling. Process Saf. Environ. Prot. 117, 639–650.
Tanaka, S., Majima, M., Kato, Kawabata, N., 2015. Performance validation of a hybrid
References ventilation strategy comprising longitudinal and point ventilation by a fire
experiment using a model-scale tunnel. Fire Saf. J. 71, 287–298.
Abidin, M.Z., Rusli, R., Khan, F., et al., 2018. Development of inherent safety bene- Tanjin Amin, Md., Khan, Faisal, Imtiaz, Syed, 2018. Dynamic availability assessment
fits index to analyse the impact of inherent safety implementation. Process Saf. of safety critical systems using a dynamic Bayesian network. Reliab. Eng. Syst.
Environ. Prot. 117, 454–472. Saf. 178, 108–117.
Barbato, Cascetta, F., Musto, M., Rotondo, G., 2014. Fire safety investigation for road Ura, F., Kawabata, N., Tanaka, F., 2014. Characteristics of smoke extraction by natural
tunnel ventilation systems-an overview. Tunnelling Undergr. Space Technol. 43, ventilation during a fire in a shallow urban road tunnel with roof openings. Fire
253–265. Saf. J. 67, 96–106.
Cong, H.Y., Wang, X.S., Zhu, P., Jiang, T.H., Shi, X.J., 2017. Improvement in smoke Wang, Y.F., Qin, T., Sun, Z.F., et al., 2016a. Full-scale fire experiments and simulation
extraction efficiency by natural ventilation through a board-coupled shaft dur- of tunnel with vertical shafts. Appl. Therm. Eng. 105, 243–255.
ing tunnel fires. Appl. Therm. Eng. 118, 127–137. Wang, K., Jiang, S., Ma, X., et al., 2016b. Numerical simulation and application study
Costantino, Musto, M., Rotondo, G., Zullo, A., 2014. Numerical analysis for reduced on a remote emergency rescue system during a belt fire in coal mines. Nat.
scale road tunnel model equipped with axial jet fan ventilation system. Energy Hazards 84, 1463–1485.
Proc. 45, 1146–1154. Wang, S., Jiang, Ma, X., et al., 2016c. Information fusion of plume control and person-
Ding, L., Khan, F., Abbassi, R., Ji, J., 2019. FSEM: an approach to model contribution of nel escape during the emergency rescue of external-caused fire in a coal mine.
synergistic effect of fires for domino effects. Reliab. Eng. Syst. Saf. 189, 271–278. Process Saf. Environ. Prot. 103, 46–59.
Ding, L., Ji, J., Khan, F., 2020a. Combining uncertainty reasoning and deterministic Wang, Y.F., Sun, X.F., Li, B., et al., 2017a. Small-scale experimental and theoretical
modeling for risk analysis of fire-induced domino effects. Saf. Sci. 129, 104802. analysis on maximum temperature beneath ceiling in tunnel fire with vertical
Ding, L., Khan, F., Ji, Jie, 2020b. A novel approach for domino effects modeling and shafts. Appl. Therm. Eng. 114, 537–544.
risk analysis based on synergistic effect and accident evidence. Reliab. Eng. Syst. Weng, M., Yu, L., Liu, F., et al., 2014. Full-scale experiment and CFD simulation on
Saf., http://dx.doi.org/10.1016/j.ress.2020.107109. smoke movement and smoke control in a metro tunnel with one opening portal.
Gannouni, Soufien, Ben, Maad Rejeb, 2015. Numerical study of the effect of blockage Tunnelling Undergr. Space Technol. 42, 96–104.
on critical velocity and back-layering length in longitudinally ventilated tunnel Weng, M.C., Lu, X.L., Liu, F., Du, C.X., 2016. Study on the critical velocity in a sloping
fires. Tunnelling Undergr. Space Technol. 48, 147–155. tunnel fire under longitudinal ventilation. Appl. Therm. Eng. 94, 422–434.
Gao, R., Li, A., Zhang, Y., et al., 2015. How domes improve fire safety in subway Yang, Khan, F., Yang, M., et al., 2018. A numerical fire simulation approach for effec-
stations. Saf. Sci. 80, 94–104. tiveness analysis of fire safety measures in floating liquefied natural gas facilities.
Gao, Z.H., Liu, Z.X., Wan, H.X., Zhu, J.P., 2016. Experimental study on longitudinal Ocean. Eng. 157, 219–233.
and transverse temperature distribution of sidewall confined ceiling jet plume. Yazdi, M., Khan, F., Abbassi, R., Rusli, R., 2020. Improved DEMATEL methodology for
Appl. Therm. Eng. 107, 583–590. effective safety management decision-making. Saf. Sci. 127, 104705.
Guo, X., Pan, X., Wang, Z., et al., 2018. Numerical simulation of fire smoke in extra- Yuan, L., Smith, A.C., 2015. Numerical modeling of water spray suppression of con-
long river-crossing subway tunnels. Tunnelling Undergr. Space Technol. 82, veyor belt fires in a large-scale tunnel. Process Saf. Environ. Prot. 95, 93–101.
82–98. Yuan, L., Thomas, R.A., Rowland, J.H., et al., 2018. Early fire detection for underground
He, B., Jiang, X., Yang, G., et al., 2017. A numerical simulation study on the forma- diesel fuel storage areas. Process Saf. Environ. Prot. 119, 69–74.
tion and dispersion of flammable vapor cloud in underground confined space. Zhong, Shi, C., He, L., et al., 2016. Smoke development in full-scale sloped long
Process Saf. Environ. Prot. 107, 1–11. and large curved tunnel fires under natural ventilation. Appl. Therm. Eng. 108,
Ingason, H., 2008. State of the art of tunnel fire research. Fire Saf. Sci. 9, 33–48. 857–865.
Ji, Jie, Guo, Fangyi, Gao, Zihe, Zhu, Jiping, Jinhua, Sun, 2017. Numerical investiga- Zhou, T.N., Liu, J.H., Chen, Q.P., Chen, X., Wang, J., 2017. Characteristics of smoke
tion on the effect of ambient pressure on smoke movement and temperature movement with forced ventilation by movable fan in a tunnel. Tunnelling
distribution in tunnel fires. Appl. Therm. Eng. 118, 663–669. Undergr. Space Technol. 64, 95–102.
Krol, M., Krol, A., Koper, P., Wrona, P., 2017. Full scale measurements of the opera-
tion of fire ventilation in a road tunnel. Tunnelling Undergr. Space Technol. 70,
204–213.

You might also like