Available online at www.sciencedirect.com
Building and Environment 39 (2004) 775 – 782
www.elsevier.com/locate/buildenv
Smoke control based on a solar-assisted natural ventilation system Wenting Dinga;∗ , Yoshikazu Minegishia , Yuji Hasemib , Tokiyoshi Yamadac a Graduate
School of Science and Engineering, Waseda University, Okubo-3, Shinjuku-ku, Tokyo 169-8555, Japan of Architecture, Waseda University, Okubo-3, Shinjuku-ku, Tokyo 169-8555, Japan c National Research Institute of Fire and Disaster, Nakahara 3 chome, Mitaka, Tokyo 181-8633, Japan b Department
Received 30 June 2003; received in revised form 9 December 2003; accepted 7 January 2004
Abstract The possibility of using the same system for natural ventilation and smoke control is examined. As an example, a prototype building is proposed. The proposed prototype is an eight-storey building with a solar chimney on top of the atrium. Reduced scale model experiments and Computational 7uid dynamics analysis are conducted. As a result, when the area ratio of outlet to inlet is greater than 2, air change rate of the utility space reaches over 2 times per hour and in an event of a 8re breaking out in the atrium, the neutral pressure plane of the smoke layer stays inside the chimney. Smoke in8ltration into adjacent spaces used for evacuation is prevented. ? 2004 Elsevier Ltd. All rights reserved. Keywords: Natural ventilation; Smoke control; Solar chimney; CFD; Model experiment
1. Introduction Recently with people paying more attention to energy conservation and indoor air quality, natural ventilation is widely adopted in modern buildings, among which the plan of utilizing atria or vertical shafts to promote stack ventilation is not scarce. To improve natural ventilation e=ectiveness, the boundary of atrium and adjacent spaces usually is completely open; this runs contrary to the traditional ways of designing 8re safety buildings with horizontal compartmentation and vertical separation of spaces. Smoke can travel from atrium to adjacent spaces and rooms that would be una=ected in the absence of an atrium. In previous researches, ‘sterile tube’ solution has frequently been used, but due to many restrictions on the atrium design and use of the building/atrium, it is always not favoured by designers. As for the other solutions, such as using a great number of shutters to separate adjacent spaces from the atrium, it is diAcult to ensure that all shutters work properly, because they are only used in case of emergency and if only one of those shutters fails to close, evacuation behaviour may be delayed or fail. In addition, shutters are always accompanied with mechanical smoke exhaust system, besides great exhaust rate is required to keep the smoke staying inside the chimney, the ∗
Corresponding author. Fax: +81-3-3209-7214. E-mail address:
[email protected] (W. Ding).
0360-1323/$ - see front matter ? 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.buildenv.2004.01.002
e=ectiveness of mechanical exhaust system in a low airtight space is doubtable. Proper smoke control method is urgent to be found out. In fact natural ventilation and smoke movement own the same nature—both are driven by stack e=ect and share the same objective of exhausting unnecessary gas (air or smoke) from buildings, which means it is possible to solve these two problems in the same way. Therefore in this paper, the possibility of using the same system both for natural ventilation and smoke control is examined. 2. Prototype building A prototype atrium building is proposed (Fig. 1). The proposed prototype is an eight-storey oAce building with an atrium in the south direction. On top of the atrium, a three-storey high solar chimney is built. Escape staircases and safety zones are suited on both sides of the atrium and separated from the atrium by walls. The utility space is completely open to the atrium. Evacuation route from the oAce rooms is directly connected to the safety zone without passing through the utility space. 2.1. Concept of natural ventilation The roof and the south-facing wall of the chimney are made of glass, which allows solar radiation passing through.
Chimney
11.25
Thermal storage wall
Atrium
A
30
Office Rooms
Utility Space
Atrium
22.5
37.5
7. 5
A
Safety Zone
Stair case
Glass
W. Ding et al. / Building and Environment 39 (2004) 775 – 782
7.5
776
Office Rooms
Stair case Safety Zone
15
22.5 37.5
15
22.5
Fig. 1. Outline of the prototype building (left: plan; right: A–A section).
Outlet Chimney
Outlet
Neutral Pressure Plane
Thermal storage Wall
Atrium
Atrium
Outside
Inlet
Inlet
Fig. 2. Concept of natural ventilation (+: positive to outside; −: negative to outside) (left: pressure di=erence distribution between the atrium/chimney and outside).
Fig. 3. Concept of smoke control (+: positive to outside; −: negative to outside) (left: pressure di=erence distribution between the atrium/chimney and outside).
The thermal storage wall absorbs solar radiation and warms up air inside the chimney. By stack e=ect the outside air 7ows into the atrium from the inlets (for the simplicity of the discussion, inlets are considered to assemble in the lower part), passes through the utility space/atrium and 8nally is discharged from the outlets. Temperature rise of air mainly occurs in the chimney and thus the neutral pressure plane almost automatically stays in the chimney. E=ective ventilation can be expected (Fig. 2).
(doors or gaps) between the atrium and the staircases/safety zones, smoke propagation into the staircases/safety zones can be prevented (Fig. 3). For the natural ventilation system, the ventilation e=ectiveness greatly depends on the temperature rise of the thermal storage wall; while for the smoke control system, the 7oor planning for evacuation is quite important. The prototype building just stands for one typical example.
2.2. Concept of smoke control
3. Research methods
When a 8re breaks out in any oAce room or in the atrium, smoke leaks into the atrium space. It is accumulated in the chimney space and at the same time exhausted from the outlets. Usually it is quite diAcult to keep the smoke stay within the chimney, it is predicted that the smoke layer will descend into the atrium. However, as long as the neutral pressure plane of the smoke layer can be kept within the chimney, the pressure of the atrium will be lower than that of the staircases/safety zones. Thus, even if there are openings
Scaled-down model experiments and Computational 7uid dynamics (CFD) analysis are carried out to con8rm the technical possibility of using the same system both for natural ventilation and smoke control in the prototype building. Between the scaled-down model and the full-scale prototype building, the following scale modelling are taken into account. When considering the smoke movement, the Froude number is most commonly used. It can be thought of as the ratio
W. Ding et al. / Building and Environment 39 (2004) 775 – 782
777
of the inertia force to buoyancy Fr =
U2 ; gl
(1)
where U is characteristic velocity (m/s), g is acceleration due to gravity (m=s2 ) and l is characteristic length (m). Considering the same fuel and the same ambient environment, the following relations can be obtained: 1=2 Um lm = ; (2) UF lF MTm = MTF ; Qm = QF
lm lF
(3)
5=2 :
(4)
When considering the free convection, the Grashof number is always used, which is the ratio of the buoyancy to viscous force. Gr =
g MTl3 ;
2
(5)
where g is acceleration due to gravity (m=s2 ), is thermal expansion coeAcient (1=K), MT is temperature rise (K), l is characteristic length (m) and is viscosity m2 =s. Actually it is almost impossible to preserve the Grashof number in the reduced scale model and the full-scale building. However it is found that even for the free convection, as long as the turbulent intensity of the 7ow is over some value of the Grashof number, the basic characteristic of the 7ow becomes independent on the Grashof number. For the natural-ventilated space, most of the 7ow region can be regarded as such state [1]. Therefore substituting the turbulent viscosity (It is proportional to the product of the local velocity and length [1,2]) for the Grashof number
t ˙ Ul;
(6)
g MTl : (7) U2 The turbulent Grashof number actually becomes in accordance with the Froude number. Considering MTm = MTF , then 1=2 Um lm = : (8) UF lF
(Gr)t ˙
Basically the similarity laws mentioned above can be used to establish relations between the scaled-down model and the full-scale prototype. 4. Description of the model experiments To simplify the discussion, natural ventilation of only the atrium and the utility space is discussed. Considering manual operability of the experimental model and e=ectiveness of the similarity laws between the scaled-down model and
Fig. 4. Picture of the experimental model. Table 1 Measurement instruments Measured items
Instruments
Type
Temperature Velocity
Thermocouple Anemometer Electric pressure transducer Multichannel carrier demodulator Temperature controller
K-type ±1:1◦ C Hot-wire ±0:1 m=s Diaphragm
Pressure di=erence
—
Accuracy
±0:05 Pa
—
the full-scale prototype, the experimental model is repro1 duced as 25 of the full-scale prototype (Fig. 4). Wall 1 is made of 1 mm thick transparent acrylic panel. Walls 2, 3 and 4 are made of 30 mm thick polystyrene boards to achieve low heat loss. Exterior walls of the atrium part are also made of 1 mm thick transparent acrylic panel to observe the smoke movement in the atrium. Panel heaters are attached to the interior surfaces of walls 1, 2, 3 and 4, on which surface temperature sensors are set and thus the surfaces’ temperatures can be controlled through a temperature controller. Along the centerline, air temperature distribution is measured with K-type thermocouples. Air velocities of the inlets and outlets are measured using a four-channel thermal anemometer. The pressure di=erences between outside and the atrium/chimney are measured using electric pressure transducers at the base of the atrium, the border of the atrium and chimney, and the top of the chimney along the centerline. All instruments are connected with a data collector to record the data every second. Table 1 lists the details of the instruments used in the experiments. Fig. 5 shows the outline of the experimental model and the arrangement of measured points.
778
W. Ding et al. / Building and Environment 39 (2004) 775 – 782
4.1. Experimental conditions for natural ventilation In fact the interior surface temperature of the solar chimney is dependent on solar radiation, which varies with days, seasons and regions. In this paper, considering the climate conditions of Tokyo and the material performance of the solar chimney, several simple temperature conditions (reproduced by panel heaters) are carried out as shown in Table 2. During the experiments, ambient temperature is 1 about 27◦ C. Taking the inlet area as about 45 of the 7oor area
Fig. 5. Outline of the experimental model and arrangement of measured points.
of the atrium, di=erent areas of outlet and inlet are tested as shown in Table 3. 4.2. Experimental conditions for smoke control As to the smoke control experiment, ethanol is used as the 8re source and put in a shallow metal dish with the diameter of 4:5 cm. The dish is put on the base of the atrium with 10 cm far away from the centerline where temperature and pressure di=erence are measured. For each case, 20 ml ethanol is burned and the time-averaged heat release rate is used as intensity of the heat source in simulation. It is calculated by following equation: V MHC q= ; (9) t where q is the time-averaged heat release rate (kW), is the density of ethanol (0:791 × 103 kg=m3 ), V is the volume of burned ethanol (20 × 10−6 m3 ), MHC is e=ective heat of combustion of ethanol (26:8 × 103 kJ=kg), t is the burning time (s). Considering a large quantity of smoke will be generated, larger openings are selected to conduct smoke control experiments, just as shown in Table 4. Since the temperature of the smoke is higher than that of the walls of the chimney, the e=ect of the walls’ temperature rises to smoke movement
Table 2 Setting of wall temperatures Case
Wall 1 (◦ C)
Wall 2 (◦ C)
Wall 3 (◦ C)
Wall 4 (◦ C)
1 2 3
35 40 40
35 40 45
35 40 45
50 55 60
Table 3 Area ratio of outlet to inlet (natural ventilation) Area ratio (outlet/inlet) 0.5 1 2
Inlet
Outlet
Model scale
Full scale (m2 )
Model scale
Full scale (m)2
6 cm × 8 cm × 2 6 cm × 8 cm × 2 6 cm × 8 cm × 2
6 6 6
4 cm × 6 cm × 2 4 cm × 12 cm × 2 4 cm × 24 cm × 2
3 6 12
Table 4 Area ratio of outlet to inlet (smoke control) Area ratio (outlet/inlet) 1 2 3
Inlet
Outlet
Model scale
Full scale (m2 )
Model scale
Full scale (m)2
6 cm × 8 cm × 2 6 cm × 8 cm × 2 6 cm × 8 cm × 2
6 6 6
4 cm × 12 cm × 2 4 cm × 24 cm × 2 4 cm × 36 cm × 2
6 12 18
W. Ding et al. / Building and Environment 39 (2004) 775 – 782
is considered small, the panel heaters are not used during the smoke control experiments.
5. Numerical modelling 5.1. Description of the CFD model The numerical simulations for this research are done using a general three-dimensional CFD model. In the simulation for natural ventilation, an indoor zero-equation model [3] is used where eddy viscosity t is given from an analytical equation without involvement of transport equations. In the simulation for smoke control, standard – model is used which is based on both a transport equation for turbulent kinetic energy and a transport equation for the dissipation of turbulent kinetic energy . In the simulation of natural ventilation, density variation caused by temperature rise is expressed using Buossinesq approximation, while in the simulation of smoke control, density is considered with the assumption of being governed only by temperature (pressure is set as constant as ambient). In all simulations, surface-to-surface radiation is considered and hot gas radiation is ignored. The upwind scheme has been used in the calculations. 5.2. Computational grid Gird independence is recommended to set the maximum 1 size of the control volumes smaller than 20 of the model size in each direction. Grid independence tests are carried out, respectively, for the scaled-down model and the full-scale model. For the scaled-down model, the results of the model with 0:03 m×0:03 m×0:03 m control volumes show almost no di=erence with the model with 0:05 m ×0:05 m ×0:05 m control volumes. For the full-scale model, the results of the model with 0:75 m×0:75 m×0:75 m control volumes show almost no di=erence with the model with 1 m × 1 m × 1 m control volumes. Therefore in following calculations, for the scaled-down model, the computational domain is basically divided into a grid of 0:03 m × 0:03 m × 0:03 m control volumes. Additional grid points are embedded in the part of chimney, near the walls, around the inlets and outlets and around the heat source to enable better resolution in these areas. For the full-scale model, the computational domain is basically divided into a grid of 0:75 m × 0:75 m × 0:75 m control volumes. Just as in the scaled-down model, additional grid points are embedded in the part of chimney, near the walls, around the inlets and outlets and around the heat source. 5.3. Boundary conditions To simulate the experiments, the following boundary conditions are used:
779
(1) Walls: In the simulation of natural ventilation, the interior surfaces’ temperatures of the chimney are set as same as that of the experiments. All the other surfaces are set as adiabatic. In the simulation of smoke control, all surfaces are set as adiabatic. Just as mentioned above, heat loss of the experimental model will be greater than the numerical model, but for natural ventilation, since temperature rise happens almost only in the chimney, it is thought that the assumption will not cause great di=erence on temperature distribution and so on. (2) Openings: Inlets and outlets are set as pressure boundaries that allow 7ow in and out of the domain. The pressure is set as the same as ambient. (3) Heat source: Klote and Mike [4] recommend design 8res of approximately 2000 and 5000 kW for atria with restricted fuel and atria with combustibles, respectively. In this paper, we select 2000 kW to conduct the experiments. The Froude number is used to decide the heat release rate of the 8re source in model experiments. Just as mentioned above, to preserve the Froude number, the relation of the heat release rate should be as described in Eq. (6). Be1 cause the scaling criterion is 25 , heat release rate of the 8re source in model experiments is expected to be about 640 W. Ethanol put in a dish with 4:5 cm diameter is used as the 8re source of the model experiments. According to the results, time-averaged heat release rate of the ethanol is about 700 W. Correspondently heat release rate in the full-scale model is set as 2187:5 kW. A cylinder volumetric heat source is used to simulate the 8re source, whose diameter is set to be 0:045 × 25 = 0:5625 m; and height is decided by McCa=rey [5]. Zf = 0:2Q2=5 ;
(10)
where Q is heat release rate of the 8re source.
6. Results 6.1. Results of the scaled-down model in case of natural ventilation Temperature distribution, pressure di=erence distribution and velocities of inlets and outlets with di=erent area conditions (Table 3) are examined, while the walls’ temperatures of the chimney keep the same (Case 2 in Table 2). (1) Temperature distribution: Fig. 6(a)–(c) left show the centerline temperature distribution of the atrium/chimney according to the results of experiments and simulation, which show relatively good coincidence with each other. With the increasing of the outlet area, temperature di=erence from bottom of the atrium to top of the solar chimney decreases. This is in accord with the increasing of volumetric air7ow rate. (2) Pressure di?erence distribution: Fig. 6(a)–(c) right show the pressure di=erence distribution between the
W. Ding et al. / Building and Environment 39 (2004) 775 – 782
1.4 1.2
1
1
Atrium
Height (m)
1.2
0.8 0.6
Chimney
1.4
Height (m)
1.6
Chimney
1.6
0.8
Atrium
780
0.6 0.4
0.4 CFD
0.2
0.2
CFD Experiment
Experiment
0
0 25
(a)
30
35
40
45
-0.3 -0.2 -0.1 0
Temperature (˚C)
25
0.1 0.2 0.3
30
(b)
Pressure difference (Pa)
35
40
45
-0.3 -0.2 -0.1 0
Temperature (˚C)
Chimney
1.6
0.1 0.2 0.3
Pressure difference (Pa)
1.4
1 0.8
Atrium
Height (m)
1.2
0.6 0.4 CFD
0.2
Expriment 0 25
(c)
30
35
40
Temperature (˚C)
45
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 Pressure difference (Pa)
Fig. 6. Temperature distribution and pressure di=erence distribution in the atrium/chimney (a) area ratio of outlet to inlet = 0.5, (b) area ratio of outlet to inlet = 1, (c) area ratio of outlet to inlet = 2.
6.2. CFD prediction using a full-scale model in case of natural ventilation
Velocity of experiment (m/s)
0.4 Voutlet Vinlet 0.3
0.2
0.1 0.1
0.2
0.3
0.4
Velocity of CFD (m/s)
Fig. 7. Velocities of CFD and experiments.
atrium/chimney and outside according to the results of simulation and experiments. The results of simulation agree with the experiments well. The position of the neutral pressure plane rises with increasing of the outlet area. (3) Volumetric airAow rate: Velocities of the inlets and outlets are measured to assess the ventilation eAciency. Fig. 7 illustrates the velocities of simulation and experiments. They match with each other well.
From the comparison of simulation and experiments, the validity of the CFD model is con8rmed. Here the CFD model is used to predict the ventilation eAciency of a full-scale model. Ventilation rate with di=erent wall temperature (Table 2, outlet 12 m2 /inlet 6 m2 ) and di=erent area conditions (Table 3, the walls’ temperatures of the chimney keep the same (Case 2 in Table 2)) is examined. Fig. 8 shows that with the increasing of area ratio of outlet to inlet, ventilation rate increases too, and when the area ratio goes up to 2, air change rate of the atrium/chimney reaches 2 times per hour, which means good indoor environment can be obtained. Fig. 9 shows with the same area conditions (outlet 12 m2 /inlet 6 m2 ), ventilation rate goes up with the increasing of the walls’ temperatures. In fact, the walls’ temperatures of the solar chimney depend on the quantity of solar radiation that can reach the walls. 6.3. Results of the scaled-down model in case of smoke control Simulation for the scaled-down model is conducted when the area ratio of outlet to inlet is 1(outlet 96 =inlet96 cm2 ). In experiments, temperature distribution of the upper part tends
W. Ding et al. / Building and Environment 39 (2004) 775 – 782
to become constant several minutes later. The data used below are obtained when the quasi-steady-state condition of the experiments is reached. (1) Temperature distribution: Fig. 10 shows temperature distribution in the atrium/chimney. Since in the simulation
3
Ventilation rate (m /h)
40000 2 30000 20000
1
air change rate (/h)
3
50000
10000 0
0 1
2 Case
3
Fig. 8. Variation of ventilation rate with di=erent area ratio of outlet to inlet.
3
2 30000 20000
1
10000 0
0 0.5
1
model, all walls are considered to be adiabatic, while in experiments, heat loss from the model is inevitable, which leads to the higher temperature in simulation than that of the experiments. According to the temperature distribution, it can be inferred that smoke descends to near the base of the atrium. Even with increasing the outlet area, the same phenomenon is observed. In order to prevent smoke propagating into escape staircases and safety zones, it is necessary to take measures to make the pressure of the atrium negative to that of the staircases/safety zones. (2) Pressure di?erence: Fig. 11 illustrates pressure di=erence between the atrium/chimney and outside. The neutral pressure plane is at the height of about 0:9 m, which means smoke may propagate to the upper 7oors above sixth 7oor. (3) Velocity: Fig. 12a is velocities of experiments and CFD. The velocities of CFD are slightly greater than those of the experiments. This can be explained by the low air tightness of the experimental model. Fig. 12b is the section of velocity vector across the heat source. In the chimney slight air7ow circulation can be observed, which leads to uniformly distributed temperature in the chimney. 6.4. CFD predictions using a full-scale model in case of smoke control
Air change rate (/h)
40000
3
Ventilation rate (m /h)
50000
2
Area ratio of outlet to inlet Fig. 9. Variation of ventilation rate with di=erent temperature conditions.
Fig. 13 is the pressure di=erence distribution between the atrium/chimney and outside with di=erent area ratio. When the inlet area is 6 m2 and the outlet area is 12 m2 , the neutral pressure plane is at the border of the atrium and the chimney. When the outlet area increases to 18 m2 , the neutral pressure plane goes up to in the chimney. Although for the easy construction of the experimental model, staircases and safety zones are not reproduced. As long as the pressure of the atrium is lower than that of the outside, the outside air will 7ow from the outside into the atrium; and according to the
CFD Experiment
1.4
Outlet
Chimney
Outlet
1.6
781
1 0.8
Atrium
Height (m)
1.2
0.6 0.4 0.2 0
Heat source
0
20 40 60 Temperature (˚C)
Inlet
80
Fig. 10. Temperature distributions in the atrium/chimney in an event of a 8re (left: centerline; right: Section).
782
W. Ding et al. / Building and Environment 39 (2004) 775 – 782
Chimney
30 Oulet18m 2 /Inlet6m 2 25
0.8
Height (m)
1
Atrium
Height (m)
1.2
35
0.6 0.4
20 15 10
Oulet6m 2 /Inlet6m 2
5
Oulet12m2 /Inlet6m 2
CFD
0.2
Experiment 0 -1
-0.5 0 0.5 Pressure difference (Pa)
0
1
-50
Fig. 11. Pressure di=erence distributions between the atrium/chimney and outside.
-25 0 25 Pressure difference (Pa)
50
Fig. 13. Pressure di=erence distributions between the atrium/chimney and outside.
7. Conclusions
1.1 Velocity of Experiment (m/s)
Atrium
1.4
Chimney
40
1.6
1 Voutlet1
Voutlet
0.9 0.8 0.7 Vinlet1 Vinlet 0.6 0.6
(a)
0.7 0.8 0.9 1 Velocity of CFD (m/s)
Outlet
1.1
Outlet
In this paper, through model experiments and CFD predictions, the possibility of using the same system for both natural ventilation and smoke control is con8rmed. The focus is put on controlling the position of the neutral pressure plane, which is greatly a=ected by the area ratio of outlet to inlet. To the prototype building, when area ratio of the outlet and inlet is greater than 2, desirable ventilation rate can be obtained and at the same time, in an event of a 8re, the neutral pressure plane of the smoke layer stays in the chimney and smoke spread from the atrium to staircases/safety zones can be prevented. Acknowledgements This research was carried out with the support of the JSPS Science Promotion Fund No.13450241—Control of Stack E=ect as a Basic of Fire-safety Green Buildings. The experiments were conducted with the cooperation of Satoshi Nishimoto, Satoshi Tanaka and Yusuke Egawa of Waseda University. The authors wish to thank them. References
Heat source
Inlet
(b) Fig. 12. (a) Velocities of CFD and experiments and (b) section of velocity vector.
pressure balance, the pressure of the staircases/safety zones will be between the atrium and the outside. That means the pressure of the atrium naturally becomes lower than that of the staircases/safety zones, thus smoke penetration from the atrium to the staircases/safety zones will be prevented.
[1] Takashi Shoda, Takao Tsuchiya. Modeling criteria for the room air motion, Part 1. Practical similarity criteria for the room air motion. Journal of SHASE 1981;17:1–11. [2] Shuzo Murakami, Model experiments on the natural ventilation, summaries of the technical papers of annual meeting. Architectural Institute of Japan, August 1980, p. 5–6. [3] Qingyan Chen, Weiran Xu. A zero-equation turbulence model for indoor air7ow simulation. Energy and Buildings 1998;28(2):137–44. [4] Klote JH, Mike JA. Design of smoke management systems, Atlanta: ASHRAE, Inc.; 1992. [5] McCa=rey BJ. Purely buoyant di=usion 7ames: some experimental results. NBSIR 79-1910.
