SPREAD AND PROPAGATION OF GENERIC SHOPPING MALL …

Proceedings on International Conference on Disaster Risk Management,

Dhaka, Bangladesh, January 12-14, 2019

Page | 641

SPREAD AND PROPAGATION OF GENERIC SHOPPING MALL FIRE OF

BANGLADESH UNDER DIFFERENT SCENARIOS

M. A. M. S. Shoshe1 and M. A. Rahman2

ABSTRACT

Devastating loss of property and lives due to shopping mall fire incidences is a recurring hazard in context of

Bangladesh. The present study investigates the spread and propagation of such generic shopping mall fire

with the use of Fire Dynamics Simulator (FDS) at different fire scenarios. The simulations are carried out for

two typical shopping mall designs with 246 m2 and 315 m2 floor area, for two fire sources, with two different

fuels of different heat release rates (HRR) in actual and grid system fuel distribution, a total of seven different

fire scenarios. The spread of fire, propagation of the buoyancy driven fire induced smoke, temperature profile

and carbon monoxide concentration along the shops, hallways and nearby stairs are analyzed. The

propagation of fire and smoke dispersion are found to be highly depended on the fire source position, fuel

type, fuel distribution and design geometry of the mall. Results shows, the standard grid system fuel

distribution can delay the maximum temperature and lethal CO concentration for about 100s.

Introduction

In the recent past, there are many instances of fire in the shopping malls of Dhaka city, causing significant

damage of property and loss of lives. Basundhora city, the second largest shopping mall in Bangladesh,

witnessed two fire incidents in March 2009 and again in August 2016, Zahur (2009), Rabbi (2016). Dhaka

North City Corporation (DNCC) market faced a devastating fire on January 2017, which destroyed 296 shops,

a portion of the market was collapsed on impact, and a loss of nearly Taka 200 crore, Hasan (2017). These

statistics are just the tip of the iceberg, according to Bangladesh Fire Service and Civil Defense, there were

as many as 18,048 fire incidents from July 2016 to June 2017, causing an estimated loss of Taka 430 crore,

BSS (2017). There are already 3000 shopping malls in Dhaka city and the number is increasing continuously,

Rahman (2010). The fuel distribution in shops of any typical shopping malls in Bangladesh, is even more

alarming, as typically these shops have compact stacking of ordinary combustibles, thus, high density of fuel

loads on an event of fire. Researchers are giving more attention to fire safety in accordance to the recent fire

events. The use of Computational Fluid Dynamics (CFD) to effectively simulate fire would be beneficial and

less costly than conducting a large-scale fire testing for planning of fire safety on future designs or

modification on existing ones. Ryder (2004) and Ryder (2006) investigated Fire Dynamics Simulator (FDS),

an LES code developed by the National Institute of Standards and Technology (NIST), for several validation

runs. They suggested FDS as more effective in simulating fire and smoke dispersion, than traditional CFD

codes, and also reported that, FDS required less computational power. Shen (2008) validated FDS in a hotel

arson fire and Wahlqvist (2013) validated FDS in much sophisticated well-confined mechanically ventilated

fire scenarios with the likes of nuclear industry.

Due to this wide spread acceptability of FDS software, researchers are now highly depended on FDS to

investigate the fire propagation and smoke dispersion in large open spaces like shopping mall, subway atrium

and underground tunnels. Khan (2017) with Fire Dynamic Simulator with an evacuation software and

predicted the total evacuation time for seven fire scenarios with 100 evacuees in a single storied 64 m2 floor

area shopping mall consisting seven shops. They reported a 390% increase in the total evacuation time (TET)

for a 50% increase in soot yield, thus concluded that the evacuation is highly affected by the fuel’s thermal

and reaction properties. Albis (2015) simulated four scenarios of fire for a two storied shopping mall of 50,753

m2 floor area in each floor having 144 stores with FDS. They reported that the smoke propagated to the second

floor at approximately 180s and covered the total floor within 900s after ignition. Gao (2012) with large eddy

simulator (LES) studied the dispersion of fire induced smoke in a subway atrium for six fire scenarios. They

reported, within 200s of ignition the atrium was filled with smoke and mechanical ventilation was found

useful to disperse the smoke effectively. To reduce the horizontal dispersion of smoke roof windows were

suggested.

1Assistant Professor, Dept. of Mechanical and Production Engineering, AUST, Dhaka-1208, Bangladesh 2Associate Professor, Dept. of Mechanical Engineering, BUET, Dhaka-1000, Bangladesh

Email of Corresponding Author –[email protected], [email protected]



Page | 642

Figure 1. Typical shops of Co-Operative market and MultiPlan Shopping Complex (First floor).

Figure 2. PyroSim models of Co Operative market and Multiplan Shopping Complex.

Ji (2015) proposed an empirical correlation for predicting the maximum upstream temperature in inclined

urban road tunnel fires with FDS. They reported the dimensionless upstream maximum temperature had a

nonlinear and non-monotonous relationship with tunnel slope. Although FDS was used to simulate different

fire situations as mentioned above, the literatures in context of Bangladesh were limited, and specific attention

to the high fuel density usually seen in typical shops, was absent. This study thus, investigates two typical

shopping malls of Dhaka, with high density of fuel loads in each shop, by FDS. Emphasis was given to the

nature of propagation of fire and dispersion of smoke in those shopping malls, specifically at staircase, with

the fire source, fuel type and fuel arrangement variations.

Simulation Method

The physical model

The shopping malls investigated in this study were Co Operative market at Mirpur and Multiplan Shopping

Complex in Elephant Road, Dhaka, Bangladesh. The investigated shopping malls were visited and photos

had been taken with permissions from the shop owners to replicate the shop interiors, as shown in Figure. 1.

The FDS software used in this study was a CFD model of fire-driven fluid flow. It solves Navier-Stokes

equations for low-speed, thermally-driven flow and gives emphasis on smoke and heat transport from fires.

The FDS code treats turbulence by Large Eddy Simulation (LES). The FDS codes were generated by

PyroSim, a graphical user interface for FDS. The physical models were designed in accordance to the photos

taken. Figure. 2 shows the PyroSim model for Co Operative market and Multiplan shopping complex. The

Co Operative market was designed for two types of shops, clothing stores and furniture stores. The legend

CS1 stands for the fire source at shop 1 with clothing, CS2 for fire source at shop 2 with clothing, FS1 for

fire source at shop 1 with furniture and FS2 for fire source at shop 2 with furniture in Co Operative market.

MS1 stands for the fire source shop at MultiPlan shopping complex CH1, CH2, MH1 and MH2 stands for

the hallways in those shopping malls. The zoomed portion in Figure. 3 shows the mesh refinement in near

fire regions than the other regions.

Validation and Mesh Sensitivity Study

The validation and mesh sensitivity analysis were done with an ISO-9705 compatible room from Zalok

(2007). In the FDS user guide McGrattan (2013), suggested a non-dimensional expression D*/dx to be used

to find mesh resolved case, where D* is a characteristic fire diameter as shown in Eq. 1. Here, Q̇ is the total

heat release rate of the fire, ρ∞, c∞, T∞ are the ambient density, specific heat, and temperature. McDermott



Page | 643

(2010) suggested that D*/dx =10 would be sufficient for a mesh resolved case. The validation works was

done for four D*/dx values in accordance with FDS user guide. Summery of the mesh sensitivity analysis

was given in Table 1.

Table 1. Summary of the mesh sensitivity study

Mesh D*/dx Cell Size (cm) Elements Time*

Coarse 4 28.4 768 0.044

Moderate 10 11.36 18432 1

Fine 16 7.1 64800 4

Finer 25 4.5 218700 34.67

*Dimensionless computation time, was obtained from McDermott (2010) suggested minimum D*/dx value.

Figure 3. Validation of Zalok (2007) and mesh sensitivity analysis.

The computation times were non-dimensionalized by the time required for the D*/dx =10 case, named

moderate in representations, to observe the improvement in results by cost of resources. The results of the

mesh sensitivity analysis were shown in Figure. 3. Observing the Figure. 4 and Table 1, the mesh size of

D*/dx =25, named finer, was discarded due to excessive resource uses with virtually no improvement in

results. The D*/dx =4 case, named coarse was also discarded due to its widespread data points. The D*/dx

=10 and D*/dx =16 cases, namely moderate and fine, were within acceptable range in both resolution of result

and demand of resources.

D*=(Q̇ (ρ∞

c∞T∞√g⁄ ))2

5 (1)

Simulation Details

The simulations were done with two fire sources, two types of fuel and two types of fuel arrangements in two

shopping malls. For the Co Operative market two fire sources at CS1/FS1 and CS2/FS2 were simulated as

clothing and furniture stores in actual fuel distribution observed from the shopping mall visits. The resulting

simulation cases for Co Operative market were CCS1 fire at clothing store 1, CCS2 fire at clothing store 2,

CFS1 fire at furniture store 1, CFS2 fire at clothing store 2. Then the clothing stores fuels were redistributed

to grid system distribution, a standard and less compact distribution, this results two simulation cases, CCS1G

for fire at clothing store 1 in grid system and CCS2G for fire at clothing store 2 in grid system. The MultiPlan

shopping complex was simulated for fire at MS1 for clothing store only and named MCS1. The present study

used D*/dx =15 near fire source and D*/dx =7.5 in other regions. The mesh elements for the CCS1, CFS1

and CCS1G case were 362,040, for CCS2, CFS2 and CCS2G case were 419,300, and for MCS1 case were

523,560. In FDS, the parameters for fabric were used from Zalok (2007), the heat release rate per unit area

HRRPUA was taken to be 1528 kW/m2 and CO yield as 0.0369. For furniture the default red oak wood

parameters available in PyroSim was used. The end of the hallways and staircase celling were designed as

open boundary. The investigated different fuel types and distribution in PyroSim are shown in details in

Figure. 4.

Results and Discussion

The simulations of the present study were conducted for 600s (10 minutes) to observe the flame propagation



Page | 644

and smoke dispersion in PyroSim. The fire was started with a fixed temperature igniter of one mesh element

size. Ignition occurred after the self-ignition temperature of the fuel was reached and the fire was then

propagated to other areas, measures were taken to ensure multiple flame fronts were not originated from the

igniter itself. The Figure. 5 shows the fire scenarios for CCS1 and CFS1 case for 150s, 300s, 450s and 600s.

From the Figure. 5 it was observed that the flame was propagated much rapidly with for the furniture store

than clothing store.

Figure 4. Fuel distributions of CCS1/CCS2, CCS1G/CCS2G, CFS1/CFS2 and MCS1 (from left to right) .

Figure 5. Fire and smoke propagation for CCS1(left) and CFS1(right) for 150s, 300s, 450s and 600s.

Figure 6. Fire and smoke propagation for CCS1G(left) and CCS2(right) for 150s, 300s, 450s and 600s.

Figure 7. Temperature at staircase celling in function of time for fire source at CS1/FS1(left) and at



Page | 645

CS2/FS2(right).

The intensity of the furniture store fire was already diminishing at 600s with its peak at around 450s, but the

clothing store fire was still burning at 600s and would engulfed the total floor space if more time was given.

The smoke of the furniture store fire was confined in fire source surroundings and never fully covered the

total floor, where the clothing store fire fully occupied the floor area around 540s. Figure. 6 represents the

fire scenarios for CCS1G and CCS2 cases. Changing the fuel distribution and stacking compactness in

CCS1G clearly improved the fire situation as the intensity of flame and smoke dispersion both were smaller

than the actual distribution CCS1 case.

Figure 8. CO concentration in function of time for fire at CS1/FS1(left), and at CS2/FS2(right).

Figure 9. Temperature at CH1 hallway celling in function of time at 3m(left) and 6m(right) from CS1/FS1.

Figure 10. Comparison of hallway temperature profile at 450s and 600s for CCS1(left) and CCS1G (right).

Changing the fire source to a different location in CCS2 delays the dispersion of smoke and flame propagation

to staircase due to the design of geometry but had the same intensity and propagation as of CCS1. Figure. 7

shows the temperature at staircase celling in function of time for the fire at shop 1 case, i.e., CCS1, CFS1 and



Page | 646

CCS1G at left and for the fire at shop 2 case, i.e., CCS2, CFS2 and CCS2G at right. The furniture fuel case

for CFS1 had reached maximum temperature of around 160s much faster than the CCS1 and CCS1G cases.

The grid system of CCS1G clearly delayed the temperature increase at staircase for 100s than the actual CCS1

case. As the fire source was moved at shop 2, the flame had minimum influence on the temperature at staircase

due to geometric restrictions. Figure. 8 represents the CO concentration at staircase celling for fire at shop 1

case, i.e., CCS1, CFS1 and CCS1G at left. The red dashed line shows 50 ppm, the lethal dose for CO

concentration. As the furniture store had rapid flame movement 50 ppm threshold exceeded for furniture fire

much earlier than that of clothing fire at around 160s and continued to be higher than the threshold for the

entire simulation time. For the clothing store CCS1 at around 400s the lethal concentration of CO exceeded,

CCS1G here also delayed the CO lethal range for around 100s. For the cases fire at shop 2 the design geometry

again played important role to restrict the CO concentration under the lethal threshold at staircase celling for

the entire simulation time.

Figure 11. Fire and smoke propagation for MCS1 model for 150s, 300s, 450s and 600s (left) Temperature

and CO concentration in function of time at staircase for MCS1 case (right)

Figure. 9 shows the temperature profile at celling for the CH1 hallway at 3m(left) and 6m(right) horizontal

distance from the fire source for shop 1 cases. The temperature distribution portrays that, for furniture fire,

the temperature of the entire hallway reached around 400oC just 150s after ignition, and gradually declined

to 130oC at 600s. For clothing fire, the temperature reached around 100oC at 360s and jumped suddenly to

400oC due to flash over at 450s. The grid distribution again delayed the fire propagation for 100s and the

sudden flash over was absent in this scenario. On Figure. 10 the temperature profile at the middle of CH1

hallway was shown, for CCS1(left) and CCS1G(right). These temperature profiles further suggest that using

a grid system distribution and less compact stacking of the combustibles reduced the intensity of the fire.

From the Figure. 8, 9 and 10 it was concluded that, the furniture fire was rapid and the temperature and CO

concentration breached the lethal threshold as hallways was fully engulfed by flame in just under 3 minutes

of ignition. For the clothing fire the fire engulfed the hallway just under 8 minutes despite reducing the fuel

load to more than half of the actual load in grid distribution. On left of Figure. 11 the fire scenario of MCS1

was portrayed. The large open space in the middle aids the smoke to disperse more horizontally rather than

concentrate on hallways thus delaying the loss of visibility and CO concentration threshold. The right of

Figure. 11 shows the temperature and CO concentration for MCS1 case at staircase, the increase temperature

and CO concentration found to be delayed due to the large open space as also observed from the flame and

smoke dispersion from left of Figure. 11. Despite the large open space the MCS1 fire too become deadly in

about 8 minutes. Observing from these results it can be suggested that, the flame propagation and smoke

dispersion depends highly on the fuel type and fuel distribution at fire source, also with the placement of

source with respect to the design geometry of the shopping mall.

Conclusions

A numerical study was performed with FDS to observe the flame propagation and smoke dispersion in two

typical shopping malls with highly compact fuel distribution in context of Bangladesh. Investigations were

carried out for two shopping malls with two fire sources and two fuel types common typically in those

shopping malls. A standard grid type less compact fuel distribution was investigated and compared against

the actual fuel distribution. The findings can be summarized as follows,



Page | 647

• The propagation of flame and smoke were highly influenced by the fuel type used and fuel

distribution of the shops. The geometric position of the fire source shop was critical for the

temperature and CO concentration at staircase, a critical position for evacuation.

• The grid type fuel distribution was able to delay the lethal CO concentration threshold and maximum

temperature for a minimum of 100s for the clothing fire cases.

• The fire produced by the furniture store was rapid and fully engulfed the hallway in under 3 minutes

(180s) of ignition. And for all the clothing store cases considered, the fie engulfed the adjacent

hallway within 8 minutes (480s) of ignition. Diminishing the fuel load to more than half in grid

distribution failed to produce any eye widening results in this regard.

The present study did not consider different types of fuels in adjacent shops, a common situation in shopping

malls, this could be studied further to improve the prediction of the fire situations considered.

Acknowledgments

The PyroSim software support from ThunderHead Engineering is gratefully acknowledged.

References

Albis, K. A., Radhwi, M. N., and Gawad, A. F. A. (2015) “Fire Dynamics Simulation and Evacuation for a

Large Shopping Center (Mall): Part I, Fire Simulation Scenarios”, American Journal of Energy

Engineering, 3 (4-1), 52-71.

Bangladesh Sangbad Sangstha (2017) “Fire causes Tk 430cr loss in one year: Fire chief”, The Daily Star

online. [Accessed on July 12, 2018]

Gao, R., Li, A., Hao, X., Lei, W., and Deng, B. (2012) “Prediction of the spread of smoke in a huge transit

terminal subway station under six different fire scenarios”, Tunnelling and Underground Space

Technology, 31, 128-138.

Hasan, M. K., Rabbi, A. R. and Mahmud, A. H. (2017) “Fire partially destroys Gulshan market”, Dhaka

Tribune online. [Accessed on July 12, 2018]

Ji, J., Wan, H., Li, K., Han, J., and Sun, J. (2015) “A numerical study on upstream maximum temperature in

inclined urban road tunnel fires”, International Journal of Heat and Mass Transfer, 88, 516-526.

Khan, E. A., Ahmed, M. A., Khan, E. H., and Majumder, S. C. (2017) “Fire Emergency Evacuation

Simulation of a shopping mall using Fire Dynamic Simulator (FDS)”, Journal of Chemical

Engineering, 30 (1), 32-36.

McDermott, R. J., Forney, G. P., McGrattan, K., and Mell, W. E. (2010) “Fire Dynamics Simulator Version

6: Complex Geometry, Embedded Meshes, and Quality Assessment”, National Institute of

Standards and Technology, NIST.

McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C., and Overholt, K. (2013) “Fire

dynamics simulator, user’s guide”, NIST special publication, 1019, 41.

Rabbi, A. R. (2016) “Bashundhara mall fire under control”, Dhaka Tribune online. [Accessed on July 12,

2018]

Rahman, M. S. (2010) “The Only Solution”, Forum - A Monthly Publication of The Daily Star, 3 (3).

Ryder, N. L., Sutula, J. A., Schemel, C. F., Hamer, A. J., and Van Brunt, V. (2004) “Consequence modeling

using the fire dynamics simulator”, Journal of hazardous materials, 115 (1-3), 149-154.

Ryder, N. L., Schemel, C. F., and Jankiewicz, S. P. (2006) “Near and far field contamination modeling in a

large scale enclosure: Fire Dynamics Simulator comparisons with measured observations”, Journal

of hazardous materials, 130 (1-2), 182-186.



Page | 648

Shen, T. S., Huang, Y. H., and Chien, S. W. (2008) “Using fire dynamic simulation (FDS) to reconstruct an

arson fire scene”, Building and environment, 43 (6), 1036-1045.

Wahlqvist, J., and Van Hees, P. (2013) “Validation of FDS for large-scale well-confined mechanically

ventilated fire scenarios with emphasis on predicting ventilation system behavior”, Fire Safety

Journal, 62, 102-114.

Zahur, A.B.M.S. (2009) “Bashundhara City on fire”, The Daily Star online. [Accessed on July 12, 2018]

Zalok, E. and Hadjisophocleous, G. (2007) “Characterizing of design fires for clothing stores”, Proceedings

of the 5th International Seminar on Fire and Explosion Hazards, 23, 27.

Documents

SPREAD AND PROPAGATION OF GENERIC SHOPPING MALL …