You are on page 1of 8

http://irc.nrc-cnrc.gc.

ca

Application of numerical techniques to study fire growth and smoke movement in an atrium

NRCC-45618 Kashef, A.; Gao, L.; Bnichou, N.

A version of this document is published in / Une version de ce document se trouve dans: 2006 Combustion Institute/Canadian Section Spring Technical Meeting, Waterloo, Ontario, May 14, 2006, pp. 1-6

APPLICATION OF NUMERICAL TECHNIQUES TO STUDY FIRE GROWTH AND SMOKE MOVEMENT IN AN ATRIUM
Ahmed Kashef, Lixin Gao, and Noureddine Benichou
Fire Research Program, Institute for Research in Construction, National Research Council of Canada, Ottawa, Ontario, Canada K1A 0R6 School of Municipal and Environmental Engineering, Harbin Institute of Technology, P. R. China 150090

ABSTRACT
To evaluate the merits of the different methods of simulating fire dynamics, this paper presents a computational study of fire growth and smoke movement in a mechanically exhausted atrium. The numerical results were obtained using three approaches: simple correlations, zone modeling, and Computational Fluid Dynamics (CFD) technique. This paper compares the predictions of the various numerical models with the experimental data obtained in a series of tests conducted at the National Research Council of Canada (NRC) to investigate the effectiveness of smoke management systems in large spaces.

NOMENCLATURE
.

m: Qc : z: Tp :

mass flow rate at height z [kg/s] convective heat release rate [kW] clear height above the top of the fuel [m] average plume temperature at elevation z [C]

Ta : ambient temperature [C] C p : specific heat of plume gases [kJ/kgC] C CO2 : volumetric concentration of CO2 at a point

T : mean temperature rise above ambient [K] M : molecular weight of air M CO2 molecular weight of CO2

INTRODUCTION
An atrium within a building is a large open space created by an opening, or series of openings, in floor assemblies, thus interconnecting two or more storeys of a building. As such, an atrium violates the concept of floor-to-floor compartmentation, which is intended to limit the spread of fire and smoke from the floor of fire origin to other storeys inside a building. As a result, smoke management systems have become an indispensable part of fire protection systems in large interconnected spaces with high ceilings, such as atria and covered malls. These buildings present some of the most difficult fire protection challenges with their complex

Corresponding author. E-mail address: lixin.gao@nrc-cnrc.gc.ca, Fax: (613) 954-0483

geometry, possibility of containing large quantities of fuel, and the fact that they usually involve a large number of occupants. To demonstrate the use of these different computational tools for the fire safetyengineering appraisal of these building types, the fire and smoke movement behavior of a mechanically exhausted atrium has been examined. The analysis was conducted using predictive models that range from simple correlations through relatively simple computer-based zone models to sophisticated and computationally intensive CFD models. Simple Correlations Simple correlations are derived from experimental data and provide a simple means of calculating individual factors that collectively can be used in the design of fire protection systems. In general, they are limited to certain fire conditions (e.g. fire size, space dimensions, and fire scenarios) and are not appropriate outside the scope of these conditions. Equations 1 through 3 present a few correlations that are typically used. (a) Steady clear height with upper layer mechanical exhaust For a steady-state condition, the mass flow rate into the upper layer is given by [1]:
1 m = 0.071Qc / 3 z 5 / 3 + 0.0018Qc .

(1)

Where:
.

m = mass flow rate at a height z , kg/s; Qc = convective heat release rate, kW; z = clear height above the top of the fuel, m.
(b) Upper layer smoke temperature The average plume temperature can be estimated, from the first law of thermodynamics [2], using the following equation:

T p = Ta +

Qc
.

(2)

mCp
Where: T p = average plume temperature at elevation z , C;

Ta = ambient temperature, C; C p = specific heat of plume gases, kJ/kgC.


Equation 2 assumes that an upper layer is adiabatic due to the fact that heat transfer from the layer to atrium walls and ceiling is little. Moreover, radiative heat transfer from the smoke layer is small [2]. (c) Upper layer CO2 concentration Upper layer CO2 concentration can be calculated using the following equation [1]:
C CO2 / T = ( M / M CO2 )c p m/ Qc
.

(3)

Where:

C CO2 = volumetric concentration of CO2 at a point; T = temperature rise at the same point, K; M = molecular weight of air; M CO2 = molecular weight of CO2.
Zone Models Zone models are one-dimensional models that divide a fire compartment into a number of characteristic zones, such as the upper layer, the lower layer, the fire plume and the compartment boundary. Conditions, such as temperature, pressure, and density in each zone, are assumed to be uniform. Each zone is modeled separately and then linked together through fluid dynamic and heat transfer equations. Zone models can produce a fairly realistic simulation under most conditions [3]. The Consolidated Model of Fire and Smoke Transport, CFAST, is an example of such group of models. CFAST is a two-zone fire model used to calculate the evolving distribution of smoke, fire gases and temperature throughout compartments of a building during a fire [4]. CFD Models Due to the rapid development of computer technology, the use of CFD models to simulate fire development and smoke movement is increasing quickly. The details of fluid flow and heat transfer provided by CFD models can prove vital in analyzing problems involving far-field smoke flow, complex geometries, and impact of fixed ventilation flows. The current paper uses the Fire Dynamic Simulator (FDS) CFD model [5]. FDS solves, utilizing the Large Eddy Simulation (LES) approach, a form of high-speed filtered Navier-Stockes equations valid for a low-speed (low Mach number) fire-induced buoyancy-driven flow. These equations are discretized in space using second order central differences and in time using an explicit, second-order predictor-corrector scheme.

DESCRIPTION OF THE EXPERIMENTAL FACILITY


The experimental facility [6] was a large compartment with dimensions of 9m x 6m x 5.5m height. The interior wall surface of the compartment was insulated using 25 mm thick rock fibre insulation. A fan was used to supply fresh air into the compartment through openings in the floor around the walls. These openings had a width of 0.1 m, and a total length of 22.8 m. Thirty-two exhaust inlets, each with a diameter of 150 mm, were located at the ceiling Figure 1: Test Facility Dimensions (Figure 1), and were used to

extract hot gases during the tests. The volumetric flow rate out of these inlets was continuously measured during a test. A square propane sand burner was used for the fire source. The burner was capable of simulating fires ranging from 15 kW to 1,000 kW with three possible fire areas: 0.145 m2, 0.58 m2 and 2.32 m2. The heat release rate (HRR) of the fire was determined using two methods. The first method computes the HRR from the volumetric flow rate of propane supplied to the burner. The second method was based on the oxygen depletion method using oxygen concentrations, temperature and volumetric flow rate measured in the main exhaust duct. During the tests, the propane flow rate was adjusted to a pre-determined level to produce the required HRR, namely 150 kW in the initial stage, 250 kW in the intermediate stage, and 600 kW in the final stage of the test. The fire was maintained steady for about 15 min at each HRR to allow stable conditions to be reached in the test facility. The room was instrumented with thermocouples and pitot tubes for velocity measurements. Also, gas inlets were located in the room for extracting gas samples to determine CO2 concentrations at various locations. More details of instruments, measurements, and test procedure can be found in [6].

RESULTS AND DISCUSSION


For the CFD model, the computational domain was divided into a uniform grid of 180 x 120 x 110 control volumes (2,376,000 node). The walls of the enclosure (being insulated with smooth surfaces) were modeled as solid, adiabatic and hydro-dynamically smooth boundaries. At the ceiling vents, a constant mass flow rate was assigned based on the quasi-steady state experimental data. Floor openings were modeled as open vents to allow inlet velocity to freely develop corresponding to the mass flow rate downstream. The inlet air temperature was assumed to be 24 C. In the CFD simulations, propane was introduced over an area and at a rate corresponding to the area of the pan used and the HRR of the test being considered. The simulations followed a transient approach and continued until steady conditions were established in 6 the room. Typically, it took less than 5 Experimental Centerline: x/X=1 & y/Y=1 x/X=0.98 & y/Y=0.98 x/X=0.97 & y/Y=0.97 5 min of simulation time to obtain a x/X=0.95 & y/Y=0.95 steady solution. Each simulation 4 required a total CPU time of about 300 h. 3 Figure 2 shows predicted and 2 experimental profiles of temperature rise along the vertical centerline of the 1 room at the steady state for the 150 kW 0 HRR. The off-centerline predictions 0 200 400 600 800 1000 1200 (few centimeters away from the Temperature (o C) centerline - x/X and y/Y values less than 1), appear to be in a better Figure 2: Temperature rise at room center (HRR: 150kW) agreement with the experimental profile. At exactly the room centre, there is rather a greater difference between experimental data and predictions. This might be
Elevationt (m)

3 Figure 3 shows the predicted and experimental vertical profiles of 2 temperature rise at the one-quarter point 1 of the room for 250 kW HRR. The predicted temperature profiles at 0 0 10 20 30 40 50 different vertical locations across the Temperature ( C) modeled domain indicate that similar conditions exist throughout the room, Figure 3: Temperature rise along quarter-line supporting the assumption of zone (HRR: 250kW) models that two layers are formed in the room.
o

Elevation (m)

attributed to thermocouples being located slightly off the room centre. Similar observations were noted for the other two HRR, namely, 250 kW and 600 kW.

6 Experimental 5 Simulation

60

Figure 4 shows the predicted and experimental vertical profiles of CO2 concentration at the one-quarter point of the room for 600 kW HRR. Similarly, profiles of CO2 concentration at the one-quarter point of the room support the assumption of zone models. While temperature and CO2 6 predictions appear to be realistic, Experimental Simulation discrepancies exist between the 5 numerical and the experimental 4 profiles, especially in the hot layer. In this region, the experimental 3 temperature is uniform and the 2 interface between the hot and cold layers is clearly defined. Predicted 1 temperature and CO2 profiles show that 0 the two parameters increase with height 0 0.2 0.4 0.6 0.8 1 1.2 and do not have a layer with uniform CO Concentration (VOL %) temperature. Moreover, the predicted height of the transition layer is greater Figure 4: CO2 profile along quarter-line than the measured data. Similar (HRR: 600kW) observations were previously reported [7]. These discrepancies may be attributed to the differences in values of exhaust volumetric flow rates between experimental and numerical input data and to the thermal radiation estimated by the numerical model.
Elevation (m)
2

Comparisons with Simple Correlations, Zone Model, and CFD Model Table 1 compares experimental data with results from simple correlations, zone model, and CFD model. Comparisons are made for the interface height, upper layer temperature, and upper layer CO2 concentration. In general, CFD results seem to under-predict the interface height and CFAST results seem to over-predict this value. The correlation predictions compare well with the experimental interface height for the lower HRR, but there is a great discrepancy for the

higher HRR. For the upper layer temperature, all of the predictive results compare well with the experimental values for the lower HRR, but they are higher for the higher HRR. For the upper layer CO2 concentration, CFD and CFAST results compare well with the experimental data for lower HRR, but they are higher than the experimental data for the higher HRR, while correlation predictions are consistently higher than the experimental data. Table 1: Comparison of Predicted and Experimental Data HRR (kW) 150 250 600 150 250 600 150 250 600 Experiment (m) CFD (m) Correlation (m) Interface Heights 4.0 4.4 4.4 3.7 4.0 3.6 2.8 3.1 2.0 Upper Layer Temperature 23 24 25 46 49 47 123 170 171 Upper Layer CO2 Concentration 0.16 0.13 0.25 0.23 0.22 0.39 0.60 0.88 0.98 CFAST (m) 3.7 3.3 2.4 25 46 143 0.15 0.27 0.97

CONCLUSIONS
Simple correlations, zone model (CFAST), and CFD model (FDS Version 4) were used to evaluate fire protection systems in an atrium space. Comparisons of the CFD predictions with experimental data were performed. In general the comparisons indicated that the predicted upper layer temperatures and interface heights compare well with the experimental values. However, the temperature profiles comparisons showed some differences. While the experimental data showed that the temperature in the upper layer is uniform, the predicted temperature was not as uniform. Comparisons were also presented between results of the simple correlations and zone model. Both the results of the simple correlations and the zone model compared well with the experimental data.

REFERENCES
1. Heskestad, G., Engineering Relation for Fire Plumes, Fire Safety Journal, 7, 25:32 (1984). 2. Klote J. K., Method of Predicting Smoke Movement in Atria With Application to Smoke Management, NISTIR 5516, National Institute of Standards and Technology, Gaitherburg, MD, 1994. 3. Jones, W. W., State of the Art in zone Modeling of Fires, National Institute of Standards and Technology, Gaitherburg, MD, 2001. 4. Jones, W. W., Peacock, R. D., Forney, G. P., and Reneke, P. A. , CFAST Consolidated Model of Fire Growth and Smoke Transport (Version 6) - Technical Reference Guide, NIST

Special Publication 1026, National Institute of Standards and Technology, Gaitherburg, MD, 2005. 5. Kevin McGrattan, Fire Dynamics Simulator (Version 4) Technical Reference Guide, NIST Special Publication 1018, National Institute of Standards and Technology, Gaitherburg, MD, 2005. 6. Lougheed, G. D., Hadjisopocleous, G. V., Investigation of Atrium Smoke Exhaust Effectiveness, ASHRAE Transactions, V. 103, 1:15(1997). 7. Kashef, A., Benichou, N., Lougheed, G. D., McCartney C., A Computational and experimental study of fire growth and smoke movement in large spaces, NRCC 45201(2002).

You might also like