CFD and experimental studies of room fire growth on wall lining materials

ELSEVIER

PII:

Fire Safety Journal 27 (1996) 201-238 ~) 1997 Elsevier Science Limited Printed in Northern Ireland. All rights reserved 0379-7112/96/$15.00 S0379-7112(96)00044-6

CFD and Experimental Studies of Room Fire Growth on Wall Lining Materials Zhenghua Yan* & G/3ran Holmstedt Department of Fire Safety Engineering, Lund University, Box 118, S-221 00 Lund, Sweden (Received 8 January 1996; revised version received 29 April 1996; accepted 1 June 1996)

ABSTRACT CFD simulation and experimental tests have been carried out to study the room corner fire growth on combustible wall-lining materials. In the CFD simulation, the turbulent mass and heat transfer, and combustion were considered. The discrete transfer ( D T ) method was employed to calculate the radiation with an absorptivity and emissivity model employed to predict the radiation property of combustion products including soot, C02 and 1120, which are usually the primary radiating species in the combustion of hydrocarbon fuels. The temperature of the solid boundary was determined by numerical solution of the heat conduction equation. A simple and practical pyrolysis model was developed to describe the response of the solid fuel. This pyrolysis model was first tested against the Cone Calorimeter data for both charring and non-charring materials under different irradiance levels and then coupled to CFD calculations. Both full and one-third scale room corner fire growths on particle board were modelled with CFD. The calculation was tested with various numbers of rays and grid sizes, showing that the present choice gives practically grid- and ray numberindependent predictions. The heat release rate, wall surface temperature, char depth, gas temperature and radiation flux are compared with experimental measurements. The results are reasonable and the comparison between prediction and experiment is fairly good and promising. ~) 1997 Elsevier Science Ltd. All rights reserved. NOTATION c~ Hc Hg

Specific heat capacity ( J / k g / K ) C o m b u s t i o n heat of the pyrolysed gas fuel (J/kg) H e a t of gasification (J/kg)

* Author to whom correspondence should be addressed. 201

202

Opy H~ mfu

m,,x M Rfu S At T

T,,

Z. Yan, G. Holmstedt

Reaction heat of the pyrolysis (J/kg) Enthalpy of the solid fuel (J/kg) Enthalpy of the pyrolysis gas fuel (J/kg) Fuel mass fraction Oxygen mass fraction Sub-divisions of the thermal node used in the pyrolysis model Fuel consumption rate (kg/mVs) Stoichiometric ratio Time-step interval (s) Temperature (K) Ambient temperature (K) Pyrolysis temperature (K)

Greek symbols o Stefan-Boltzmann constant 5.67 x l0 ~ ( W / m 2 / K 4) p Density (kg/m 3) Pcha,Density of char (kg/m 3) pv~r Density of virgin material (kg/m 3)

1 INTRODUCTION As progress is made in the understanding of the process of flame spread, many attempts have been made to develop theoretical models to predict the flame spread and fire growth, with varying levels of complexity and accomplishment. Frey and T'ien' attempted to simulate the flame spread on a thermally thin fuel layer by numerical solution. In their study the gas-phase heat conduction was assumed to be the primary mode of forward heat transfer. Di Blasi et al. 2 developed a theoretical model to study the effect of gas flow on the flame spread. By prescribing the flow field and neglecting the buoyancy, the energy and species equations were solved numerically to predict the flame spread rate dependence on the flow characteristics. A two-dimensional analytical model was presented by deRis 3 for a laminar diffusion flame spread against an air stream over the surface of both thermally thick and thin fuels. An unrealistic uniform gas velocity profile was assumed, the radiation was ignored in the thermally thin fuel case, and was treated as an exponentially decreasing form in the thermally thick fuel case. The most significant semi-empirical flame spread model was probably provided by Quintiere. 4 It was based on heat transfer and developed to analyse the one-dimensional lateral flame spread experimental results. He postulated that the heat flux from the flame is

Fire growth on wall lining materials

203

constant over a small distance ahead of the flame and zero elsewhere. Due to the complexity of the flame spread problem, the above-mentioned models were developed with varying aspects and levels of simplification. The general applicability of these models to the prediction of fire growth is quite restricted by the assumptions made. A typical flame spread scenario in fires is the room corner fire. As an increasingly important scenario of evaluating lining-material's fire performance in the realistic fire environment, such as ignitability, flame spread and heat release, it attracted great efforts from fire growth modellers. Some empirical or semi-empirical models have been presented which predict flame spread and room corner fire growth on wall linings on the basis of ignition, heat and smoke release data obtained in cone calorimeter tests. A simple theory was suggested by WickstriSm and G~ransson ~ for the prediction of R H R of surface material using Cone Calorimeter results. Cleary and Quintiere ~ presented a fire growth model which describes the room corner fire spread and the resultant heat release rate in terms of the Cone Calorimeter and LIFT test data, such as R H R and ignition time. Karlsson 7 made a significant contribution to this area. He calculated the heat release rate in the growing room corner fire using bench scale test data as input. However, this type of model can only provide very limited information (mainly RHR), and due to the empirical parameters and simplifications introduced, the general use of these models remains questionable. Field modelling based on the computational fluid dynamics methodology plays an important role in fire research. Over the last decade, field modelling has been quite successfully used in many different fire situations. With the fast growth of computer technology and the progress made in CFD technology and fire sub-models, it is expected for field models to continuously gain popularity and ultimately to be the most satisfactory treatment for flame spread and fire growth. However, the complexity of the flame spread problem presents a challenge to fire field modellers. Previously in fire field modelling, no actual consideration was given to fire growth and a user-specified or semi-specified fire was used as a source of heat and mass. Very recently, some exploratory works ~-"~ were carried out to study fire spread using CFD. This paper describes a preliminary step in the CFD simulation of flame spread and fire growth in a room corner fire. Physical sub-processes, including three-dimensional transient turbulent gas flow, combustion and heat transfer (convection, conduction and radiation) and the pyrolysing and charring of the solid fuel, are considered. In order to provide more information for comparisons, the one-third scale experiment was duplicated. Theoretical calculations and experimental measurements of the

204

Z. Yan, G. Holmstedt

temperature, heat release rate, radiation flux, charring pattern and char depth were found to be in good agreement.

2 THEORETICAL MODEL

2.1 Gas phase model The basis of field modelling of the gas phase is to describe the fire phenomena by solving a set of partial differential governing equations ~1-~3 which, in this study, include continuity, momentum, K-e, species and energy equations, together with the necessary auxiliary equations such as the gas properties and radiation equations, etc. The general form of the governing equations is

at

oxi

oxi \

Oxj/

where Fii is the effective diffusive coefficient and S~ is the source term representing the production and consumption of ~. The source term must be described before the governing equations can be solved, and may be obtained from combustion, radiation and the response of solid combustible material, etc. 2.1.1 Turbulence m o d e l

The standard K-e two-equation model, adapted to incorporate the buoyancy effect, was used to study the turbulence characteristics of the gas flow. Solving these two equations provides the quantities of the turbulence kinetic energy and its dissipation rate, which are used to calculate the effective diffusive coefficient, and in the combustion model. Readers are referred to the literature TM for details. 2.1.2 Combustion m o d e l A combustion model is required to provide the source terms for the species and energy equations. In this study, combustion was simulated by one-step chemical reaction, where complete oxidation is assumed when sufficient oxygen is available, and the local reaction rate is determined by the eddy dissipation combustion model to be the slowest of the

Fire growth on wall lining materials

205

turbulence dissipation rates of either fuel or oxygeng ~s

Rfu= --p~min( CRmfuCRm°x]s

/

(2)

2.1.3 Radiation model The discrete transfer (DT) radiation model TM is employed to present the radiation source term for the energy equation of the gas phase and the radiation flux to the solid surface. DT is one of the most popular methods used in the numerical calculation of radiation. It has the advantages of good accuracy, flexibility for complex geometries and, when appropriately coupled to CFD, its CPU time and storage requirements are acceptable for most workstations available. This method solves the radiation equation along a discrete set of directions (rays) from every element of the boundary surface. Consider a pencil of radiation passing through a general control volume defined by the CFD numerical grid. Assuming the scattering can be ignored, the following relation exists: T 4

I,+, = t r - - X e + I,(1 - at)

(3)

where I, is the intensity on entry and I,+, is the intensity on exit. e and are the corresponding emissivity and absorptivity, respectively, which will be provided by the radiation property model described in the following section. The radiation intensity is provided by the solution of the above equation and is then used to calculate the radiation energy source term for all the control volumes and the net radiation flux to the boundary surface elements, as follows:

(4) Rays

R.ox =

lw(fi.,

)an -

(5)

Rays

The boundary conditions for DT are as follows. (1) Wall boundary: suppose q - is the radiative flux leaving the wall, q+ is the radiative flux incident on the wall, and the wall is a grey Lambert

206

Z. Yan, G. Holmstedt

surface, then the intensity from the wall is _ _ + o'T~ 1 q -(1-ew) q +e~ -//7

~

(6)

//7

where w denotes the wall. (2) Open boundary: the open boundary is treated as a cool black body.

2.1.4 Radiation property model In the calculation of radiation, the absorptivity and emissivity of the gas are provided by Modak's simple model. ~7 The absorptivity of a h o m o g e n e o u s and isothermal mixture of soot, CO2 and H20 is calculated by ce = % + ol~ - a~ce~

(7)

where % is the absorptivity of CO2 and HzO approximated in a manner similar to that suggested by Hottel and co-workers: '~''') { T~ "'~5-''2° { T'~ 1''-~-''2° in which ~"= Pw/(P~ + P,)'G is the emissivity of the mixture, s, and ew are the emissivities of C02 at partial pressure P~ and H20 at partial pressure Pw, respectively. They are approximated by three parameters (partial pressure, pressure-path length and temperature) curve fits to spectral calculations of e~ and sw at 1 atm total pressure. A s ~ is the overlap correction for CO2 and H20 determined by a temperature-adjusted version of Leckner's approximation. TM In eqn (7), a~ is the soot absorptivity which is defined by 15 ( A,,K,,~I c~ = 1.0 - ~ q/'3' 1 + ] C2

(9)

/

where: ~3~ is the p e n t a g a m m a function; c2 is Planck's second constant; is the source temperature; l is the pathlength; k,,=7fv/A,,; A,, = 0.94/xm and f~ is the soot volume fraction which will be given by the soot model below. 2.1.5 Soot model Soot can contribute significantly to the radiation in the fire. Unfortunately, sooting is very complex and can be affected by a large number of factors. To date, there is no very good soot model available for building fires. In this study, as an approximation, soot was considered by assuming a constant soot conversion factor, chosen with reference to some experimental measurements. 2t'22 The soot formation rate is simply assumed to be

Fire growth on wall lining materials

207

locally proportional to either fuel supply rate or fuel consumption rate. No oxidation is considered. A n additional transport equation which can be written in the general form of eqn (1) is solved to calculate the soot mass concentration. Different conversion factors were used for different materials, 2.1% for propane 22 and 3% for the particle board. 2~ The conversion factors were reduced by half during a test run. Some, but not significant, changes in results were observed. The soot volume fraction, which is central to the radiation calculation, is simply determined from the soot mass concentration by assuming a constant soot density 23 of 1800 k g / m 3. 2.2 Solid wall model In this study, the solid wall surface is divided into many elements, according to the CFD grid generation, and the solid wall model is i m p l e m e n t e d for every element. In a fire, the combustible lining material and the walls exposed to the flame and hot gas are heated up through convection and radiation heat transfer. After a certain time, the combustible lining material will start to pyrolyse and burn. Consequently, the flame will spread. Two methods may be used to describe the reaction of the combustible material. One is to solve the heat conduction equation and use the test data, such as heat release rate from Cone Calorimeter tests, as input. In this method, the combustible material element is assumed to be ignited at its ignition temperature, and its burning behaviour is then assumed to be the same as that in the Cone test. However, in reality the burning behaviour of the solid material varies with the incident heat flux which may be quite different in the real fire from that in the cone test. Thus there is some uncertainty with respect to the choice of the heat release rate data from the Cone Calorimeter test, although it was found that with a suitable choice of the heat release rate data from Cone tests as input, it was possible to predict the total heat release rate, the gas temperature and the wall temperature in the r o o m fire quite reasonably. A n o t h e r m e t h o d is to solve the heat conduction equation and develop a pyrolysis model to describe the pyrolysis and charring of the solid fuel. By using the pyrolysis model, the effect of the transient heat flux can be considered and the heat release rate of each solid element is provided automatically rather than specified by data input. At the same time, the charring pattern and char depth can also be determined by this model. It is expected that this m e t h o d will remove the uncertainty involved in the first method. This m e t h o d is more complex and expensive to run. However, it was shown that the C P U time was only about 5% longer

Z. Yan, G. Holmstedt

208

when using the pyrolysis model presented below. From this point of view, it is recommended that the pyrolysis model be used. Both methods will be discussed in detail below. 2.2.1 Cone Calorimeter data input method As mentioned above, this method numerically solves the heat conduction equation to provide the temperature profile within the solid boundary for every solid element, and uses Cone Calorimeter test data 24 as input to describe the burning of the solid fuel. During fire spread, the heat conduction in the wall parallel to the solid face can be ignored. Only the heat conduction perpendicular to the face is important. Thus, the heat transfer in the wall is simplified to a onedimensional transient process for each boundary element. The following equation is solved:

O(pH)_ot

(OT)

Ox3 k~x

(10)

where H is the enthalpy given by fCp dT. cp and k are specific heat and conductivity, respectively. They are usually temperature dependent. One initial condition and two boundary conditions are required for the solution of the above equation.

Initial condition T'~ = TOast time step, x)

(11)

Boundary conditions 0T Inner surface • k ~

OX I ~=o

= hc(Tg - T,=.) + Rr, ux

Outer surface :T,,u,~dc=

Tambicnt

(12) (13)

where Rtlox is the heat source from radiation, Tg is the temperature of the gas close to the wall surface and he is convection heat transfer coefficient. These are given by field modelling of the gas phase. Since the wall, including the lining and the concrete part, is thick enough, we can reasonably assume that the outer surface temperature is equal to the ambient temperature. The critical surface temperature ignition criterion is adopted to predict the ignition of the combustible elements. The ignition temperature is taken from the work of Cleary and Quintiere 6 as 405°C for particle board. Once ignition occurs, the element is assumed to be comparable to the specimen tested in the Cone Calorimeter and begins to release heat

Fire growth on wall lining materials

209

25(] Full s c a l e

20{] I/3 s c a l e 150

"r

/

I O0

50

I

{}

12{)

I

240

I

36{)

I

48{}

600

Time (s) Fig. 1.

H e a t release rate inputs after ignition.

according to the Cone Calorimeter test. This heat release is converted into the gas fuel release data by dividing by the combustion heat of the gas fuel. The gas fuel is added to the gas phase and the combustion is determined by the aforementioned turbulent combustion model. In this study, an approximate average value of heat release rates obtained from Cone tests z4 at irradiation levels of 25 and 50 k W / m 2 was used as the input for the one-third scale case. In the full-scale case the fire scale is larger, and consequently the heat flux and the heat flux d e p e n d e n t heat release rate are higher. A theoretical calculation with the burning of solid fuel represented by the pyrolysis model indicates an average R H R of about 20% more than that of the one-third scale case (see Section 4.8.1 and Table 1). Thus, the Cone R H R input for the full-scale case was increased by 20%. These two heat release rate inputs are shown in Fig. 1. 2.2.2 Pyrolysis model As an alternative, and in order to consider the important effect of the heat flux on the heat release rate, a simple and practical pyrolysis model has been developed and implemented into CFD. This model is based on the numerical solution which affords the model much flexibility. It can easily be used in the complex cases such as those with transient incident heat flux and t e m p e r a t u r e - d e p e n d e n t material properties. As demonstrated later, it is generally applicable to charring and non-charring material. This pyrolysis model is very fast and describes the essential physics, in so far as is n e e d e d to predict correct mass loss and heat release rates. By using essentially the same pyrolysis model, John deRis and Z h e n g h u a

210

Z. Yan. G. Hohnstedt

Yah are now developing an 'equivalent properties' optimization program to analyse and fit the Cone Calorimeter test results. A data base of the 'equivalent properties' of the materials tested in the Cone can be created. By using the optimized equivalent properties, this pyrolysis model can be expected to be applicable to the realistic composite materials and be used as an alternative to the more complex and computer-time consuming models. 25,2~ 2.2.2.1 Model basis'. This method differs from the Cone data input m e t h o d mainly in its way of providing the heat release rate for the elements of the solid fuel. The one-dimensional transient heat conduction equation is also solved numerically here, but with pyrolysis and charring included. The heat conduction equation can now be written as

.(pH) + m"'(Hpy + H) +

-

/ )

(14)

where rh" -

Op i~t

-

Orb" Bx

>-0

representing the mass loss-rate of the pyrolysing material per unit volume. The third term is the energy required to heat the vaporized gas as it flows to the solid surface. This term will be zero for non-charring material and has no important effect in this study, and thus it is ignored here (but it can be very easily included). Hpy is the heat of reaction of the pyrolysis process, and can be calculated by the difference in total enthalpy of virgin material and volatile products, i.e. * * * __ * Hpy ~___ Hv,,, ,;.- Hvir.-t.p = Hv,,,.,i. Hv,.ll, +

(15)

It is worth pointing out that Hpy is a material constant and is different from the heat of gasification: H, =O°~,lrh,,,,,,,- [ h e ( L - T,=,,) + R,,ox]

(16)

mlola[

which is a local and transient value and changes considerably during the pyrolysis processF 7 For the thermally thick vaporizing material, at steady state, /i"

• r,.Hg = Hv,,,

*

Hvir

-/;,

= Hpv. +

f/

Cp dt

(17)

211

Fire growth on wall lining materials

Equation (14) can be rewritten as pep - - + t h " H p y /~t

- -- c~x

(18)

k

The material will start to pyrolyse only when its temperature reaches the pyrolysis temperature, To, and it will then keep this temperature until completely pyrolysed. Thus we have pCpaT Ot -

k

when

T
(19)

th"H..,,~ =

k

when

T -> Tp and p >

(20)

Pch,,r

The following is the detail of the numerical solution of eqn (18), which can be discretized as pCp

T,,- T;, (T,,~!- T,, At 6x +/4/~'xHpy = k \ 8x

T,,--T,,_.) fix /

(21)

where the prime indicates the previous time step and rhea, is the mass loss rate per unit area of the 6x thick strip. If we define K k Ac fix' Aw 6x (22) Do = A c + Aw + pcp(~x/At

(23)

S, = pcp6xT'o/ At - rh'~'~Hpy

(24)

ApT,, = A~.T,,+, + AwT,, , + S~

(25)

then eqn (21) becomes

Since the conductivity, k, is generally a function of temperature, it is consequently a function of x and it is not necessary for A~ to be equal to Aw.

It was found that, as demonstrated in Fig. 3, in order to obtain a reasonable result for the mass loss rate, a very fine grid was required. However, this very fine grid is unnecessary and very expensive for the temperature solution. This inconsistency is overcome by defining a reasonably coarser grid for temperature solution and refining the grid into a second grid to determine the mass loss rate, as shown in Fig. 2.

212

Z. Yan, G. Holmstedt

T1 1"2

It I

T5

I

Char layer (it charring material)

Pyrolysing zone

i Fig. 2.

Virgin material

)

i

I

Temperature solution node and grid refinement (example with N = 5, M = 10).

The temperature of the refined node, m, of the coarser node, n [we will denote this node as node (n, m) later], T,,.,, is obtained by interpolation, as shown in Fig. 2, assuming a linear distribution between T,, and T,,+I. From eqn (25), for an arbitrary small node (n, m), the energy available for pyrolysis can be approximated as H,..... = max[0.0,Ap(T, .....

- Tp)IM]

(26)

On the other hand, the mass of the volatizable material remaining in the node (n, m) which may have been completely, partially or not at all pyrolysed, is generally given by ~X

massvo, = ~ min{p,ir - pch.,max[0.0,(p ...... - Pch,r)]}

(27)

where Pm = M p '

- ( m - 1)pch.r

-

(M - m)pvir

The mass loss rate from node (n, m) is thus finally determined by rn ..... = m i n { H , , . , , , / H p y , m a s s , , , , / A t }

(28)

The overall pyrolysis rate can be obtained by summation over all the nodes and expressed as rn = ~ m /t

/)t

......

=

~.min(H,,.,,,lllpy,mass.,,,IA 0 tz

t)t

(29)

Fire growth on wall lining materials

213

The heat release rate is represented by 0 -'- ,n/L

(30)

where Hc is the heat of combustion related to the gaseous fuel produced. Generally, during flaming combustion, it has been shown that H~ is approximately constant. 2.2.2.2 M o d e l e x a m i n a t i o n . As a basic test, the pyrolysis model was used to simulate the Cone Calorimeter tests z4"28 of both charring material (particle board) and non-charring material (PMMA). The Cone Calorimeter is a device widely used to measure the mass loss rate and heat release rate per unit area under a specific incident radiant flux. Usually, the combustible samples were tested under three different levels of radiation flux: 25, 50 and 75 k W / m 2. In the tests, the total heat flux includes the external radiation, convection, radiation feedback, and once ignition occurs, the convection and radiation heating from the flame. The external radiation, q~, is provided by the electric heater of the Cone and can be specified. Before the sample is ignited, it is cooled by the ambient air through convection heat transfer. An empirical convection heat transfer coefficient of 15W/m2/K was used to represent the convection cooling. Thus 4c = 15(T,- T,~r) before ignition qc -- 0 after ignition (convection is then in the flame flux) where ~ is the sample surface temperature. Since for the charring material ~ could be quite high, the surface radiation heat loss cannot be ignored, and is approximated by (It = eo, T~

(31)

After the sample starts to burn, it will also be heated by the flame through convection and radiation. Quintiere and Rhodes z~ estimated and measured the flame heat flux for black P M M A under different external flux levels. Their study indicated that the flame heat flux was about 25-35 k W / m z. In this study, as an approximation, 35 k W / m 2 was used for the P M M A and 25 k W / m z for the particle board. The net heat flux to the sample surface is O°~, = 0~ + On..,c - 0r - Oc

(32)

Z. Yan, G. Holmstedt

214

4.00E+05. 3.50E+05-

,-i -v m,

300E+05.

- - M =

1

2.50E+05:

--

200E+05-

- ---M=

70

150E+05 !

--

140

-M=

I0

M=

I.OOE+05 5.00E+04 O.OOE+OOJ 0

100

200

300

Time

400

500

600

(s)

Fig. 3. Influenceof M.

The above equation was used as a boundary condition for the numerical solution of eqn (18). The effect of grid refinement was first investigated, as shown in Fig. 3. In order to determine the effect of the time step, the calculation was repeated using different time step intervals. Figure 4 shows that the time step interval of about 2 s is acceptable. The Cone tests of a typical charring material, particle board and non-charring material (PMMA) were simulated, at three different incident radiation levels: 25, 50 and 7 5 k W / m 2. The model requires thermal property data such as % and k, which are very likely to be temperature

3.00E+05 2.50E+05

dt = dt =

g7"

20 I.O

2.00E+05 1.50E+05 I.OOE+05 5.00E+04 O.OOE+O0

1O0

200

300

Time Fig. 4.

400

500

(s)

Effect of time-step interval.

600

215

Fire growth on wall lining materials

dependent. In the simulation of the particle board test, due to the difficulty in finding the proper data, the required properties were crudely estimated as constants. Perhaps the pyrolysis model itself can be used to infer these properties by repeatedly running to minimize the difference between the simulation results and the Cone test measurements. Properties taken directly from literature or estimated with reference to other scattered information are given below: PMMA

po = 1190 kg/m 3 k = 2.49 × 10 7T + 1.18 x 10 -4 k W / m / K Cp = 2.374 × 10-3T + 1.1 J / g / K To = 330°C Thickness = 25 mm Hg = 2.77 kJ/g Hpy = Hg -

Cp d T = Hg -

Quintiere Quintiere Quintiere Quintiere Quintiere Quintiere

and and and and and and

Rhodes TM Rhodes 2~ Rhodes 2~ Rhodes 2~ Rhodes TM Rhodes 2~

× 10-ST + 1.1) d T ~- 2.0 kJ/g

Particle b o a r d

p,, = 670 kg/m 3 k = 0.22 W / m / K Cp = 2.5 J / g / K Tp = 317°C Thickness = 10 mm Hpy = 1.3 kJ/g Hc = 12.0 kJ/g

Tsantaridis and Ostman 24 Delichatsios and Chen 2~ Estimated Delichatsios and Chen 2~ Tsantaridis and Ostman z4 Estimated Tsantaridis and Ostman z4

The calculated R H R of particle board is compared with the Cone Calorimeter measurements in Fig. 5(a)-(c). The agreement is generally reasonable. However, at the early stages there is an apparent disparity. The disagreement is similar for all the three irradiance levels. The ignition time is underpredicted. After ignition, the calculated R H R extends to a higher peak value and then decreases much faster from the maximum. During the later stages, the prediction agrees better with the experiment, especially for the low irradiance case. One important reason for the disparity could be that the initial moisture content is not considered in the calculation. Water absorbs a considerable amount of energy during evaporation, and will thus cause some delay in ignition. Another possible reason is that, due to the limitation of the thermal property data, the thermal properties were quite crudely estimated as constants. In reality, these thermal properties may have significant dependence on the temperature. The simulation is expected to be improved with the consideration of water content and the use of temperature dependent thermal properties. This work is planned for the near future.

Z. Yan, G. Holmstedt

216

4.00E+05

4.00E+05

(a)

(b) 50 kW/m 2

25 kW/m 2 ~ , 3.0OE+05

3.00E+O5 E

~

Calculation

2.00E+05

~

1.0OE+05

1.0OE*05 0.00E+O(

2.00E+05

100

0

200 300 Time (s)

0.00E+00 : ~ 0

400

30 4.00E+05,

(c)

.

200 300 Time (s)

-i

400

(d)

25

75 kW/m 2

~" E 20

3.0OE+05 ~

v

100

2.00E+05.

'-r, e~ I

lo

N

5

¢1

.OOE+05.

O.OOE+O0

~

I~10

40

(e)

"~ 30

.

200 300 Time (s)

• Experiment Quintiere's calculation Our calculation

I00

400

500

6~)

401~ ..

°

•

~ 20

~'~ 30 .~ 20

I0

• Experiment Quintiere's calculation Our calculation

N 10

. Experiment Quintiere's calculation - - Our calculation

--

0

200 3~10 400 Time (s)

I00

200

200 400 500 Time (s)

--

600

0

1;0

2ClO 200'

400 500

600

Time (s)

Fig. 5. S i m u l a t i o n of C o n e C a l o r i m e t e r tests: (a) p a r t i c l e b o a r d , 25 k W / m 2 ; ( b ) p a r t i c l e b o a r d , 5 0 k W / m 2 ; (c) particle b o a r d , 75 k W / m 2 ; (d) P M M A , 25 k W / m 2 ; (e) P M M A , 50 k W / m 2 ; (f) P M M A , 75 k W / m 2.

Fire growth on wall lining materials

217

A comparison of calculated mass release rate of P M M A with experim e n t and Quintiere's calculation is presented in Fig. 5(d)-(f). In this prediction, all the thermal properties were found and taken directly from the reference. The agreement is excellent.

2.3 Interaction between the solid and gas phase The gas phase and the solid material exchange mass, m o m e n t u m and energy through their interface. The interaction between the gas and the solid boundary material determines the boundary conditions for both gas phase partial differential equations and the solid model equations. The solid material acts as the non-slip boundary. The fluxes of m o m e n t u m and convective energy are provided by the wall function relations. The radiation between the gas and the solid is calculated from the D T model, as described above. W h e n the solid fuel starts to pyrolyse, the amount of fuel mass which is presented by the solid model is transferred into the gas region close to the solid surface.

3 D E S C R I P T I O N OF E X P E R I M E N T S

The simulated experiments were carried out by SundstriSm,3° Andersson 31 and ourselves (our measurements are only reported in this paper for the comparison of predictions). In both the full and one-third scale experiments, three walls, doorway excepted, and the ceiling were lined with the particle board which was exposed to a propane gas burner located at the r o o m corner acting as an ignition source. In the full-scale experiment, the fire r o o m was 3.6 m long, 2.4 m wide and 2.4 m high and was constructed of lightweight concrete, with a thickness of 150 mm. At the centre of one of the 2.4 × 2.4 m walls, there is a doorway of 2.0 × 0.8 m. The gas burner with a power of 100 kW is square, with a side length of 170 mm. In the one-third scale test, the scaled fire room, with the dimension of 1.2 × 0 . 8 x 0 . 8 m , was constructed of a special high-temperature resistant concrete with the following thermo-properties: k = 1 . 7 W / m / K , p = 2750 k g / m 3 and cp = 1014 J/kg/K. The doorway of 0.67 m (high) by 0.56 m (wide) was located at the centre of one of the 0.8 × 0.8 m walls. The output o f the 70 × 70 m m square gas burner was 11 kW. We duplicated the one-third scale experiment after the CFD simulation to provide some additional data for CFD comparison. The heat release

218

Z. Yan, G. Holmstedt

(a)

/

Burner

/ Fig. 6. Configuration of fire room. The numbers indicate measurement locations: (a) configuration of fire room; (b) one-third scale experiment measurements--(r) denotes the radiation measurement; (c) full-scale experiment measurement locations.

rate was measured using an oxygen consumption calorimeter, the radiation to the walls and floor was provided by a water-cooled radiation meter, and the gas and solid fuel surface temperature history was recorded by the thermocouples. The char depth was measured afterwards by scraping the char layer from the material surface. The test configurations and the measurement points are shown in Fig. 6. The numbers inside the figures denote the measurement points, and will be used later as indices in the comparison. The full-scale experimental measurements were made by Sundstr6m. 3~' In the one-third scale test, measurements one to 11 are those of Andersson, 3j the others are experimental measurements made in this study.

4 RESULTS The experimental and CFD simulation results will be presented and compared. One advantage of CFD simulation is that it can provide much detailed information on the fire, including the local and transient gas velocity, gas temperature, fuel consumption rate, species concentration, the doorway mass in-flow and out-flow rates, solid wall temperature, burning area, char depth (when using a pyrolysis model), radiation heat

Fire growth on wall lining materials

219

(b) Above burner

No. Height

N~. Hcight

5

75 cm

35 75 cm

4

60 cm

33 65 cm

3

45 cm

31 55 cm

2

30 cm

29 40 cm

1

15 cm

1141 40 43 42

34 70 crn 32 60 cm Doon~ay

39

30 50 cm 28 30 cm

a ; ~d b, locations of vertical ga temp. measurement profile a

Gas temp. (a)

Ceiling surface temp. measurements

18 21

23 24

17 20

22

16

25

Gas temp. (b)

26

19

1,5

IRlr~

14 13

12

Burner comer

44(r)

Left side wall measurements

Right side wall measurements

38

37 45(r) 17it)

Burner comer

Burner comer

Rear wall measurements

Fig. 6.

Root measurements

(Continued.)

flux, convection heat flux and the heat release rate, etc. These results will be compared with corresponding experimental data where available. Some other results will be presented only for reference. In all the following figures, the numbers indicate the corresponding measurement locations

220

Z. Yan, G. Holmstedt

(c) Above bumer

No. Height

Doorway

55

210 cm

54

172 cm

53

157 crn

52

142 crn

51

127 cm

50

97 cm

49

67 cm

X Location of vertical gas tel p, measurement profile G a s temp. m e a s u r e m e n t s

Ceiling sudace temp. measurements

Fig. 6.

(Continued.)

shown in Fig. 6. The symbol (c) denotes using Cone data input method and (p) the pyrolysis model. The sensitivity of the solution to the gas grid, wall node, time step and ray number was studied first. In the present simulation, the compartments were divided into 17 x 21 x 18 and 20 x 21 x 22 control volumes for the one-third and full-scale cases, respectively, with much denser m e s h applied in the burner and the upper layer zones. During the numerical experiments, the mesh densities in the two important zones were almost doubled and only some minor changes were found. The combustible material was represented by slabs of 1 mm, which was found sufficient for the wall modelling. DT rays number used in this study is 4 x 16. This number was increased to 8 x 32 in a test run and the test results presented only minor variations. Different time-step intervals of 1 and 2 s were tried. Only some negligible changes were found and the time-step interval of 1 s was used in this study. The convergence was controlled by verifying the regular decay of residuals and the evolution of each calculated variable to an asymptotic limit. 4.1 Heat release rate

Figure 7 shows a comparison of the R H R measured and calculated using the Cone data input method and pyrolysis model for both full and one-third scale fires. It can be seen that both theoretical results agree with the experimental data reasonably well. The large difference between the predictions and experiment at the beginning stage is due to a delay of the measuring instrument ( R H R meter).

221

Fire growth on wall lining materials (a) •

(b) exp

~ " - - pre - c(total)

1.6

- -

1.4

• - .X- - • pre - c(inside)

~- 1.2 na- 0.8 m 0.6

A

pre-p(totai)

]l

pre - p(inside)

• 160 140

1.8

-

120 ~" 100

•A

60 I

~< ,+

- .X- - • pre - c(total)

...... -I"

e x p (bY B. Andersson) exp (this work)

,~

pre - c(inside)

+

pre - Pltotal)

&°~

&

pre - p ( i n s ~ L ~ ¢ ~

0.4 0.2 0 0

50

100 Time (s)

150

200

0

100

200

300

Time (s)

Fig. 7. Calculated and measured heat release rate. The symbol (c) denotes using Cone data input method and (p) the pyrolysis model: (a) full scale; (b) one-third scale.

There are two curves for each prediction. One is the heat release rate inside the room, and the other one represents the total heat release including that outside the fire room. These two curves are coincident with each other at the early stages of the fire. At this time, no flame comes out through the doorway. At the later stages, especially at flashover, the difference between them becomes greater, indicating that some fuel has left the fire room through the opening. When the fire is close to the flashover, the predicted flame spread and fire growth are considerably slower than that obtained experimentally. Although the calculated R H R increase is accelerated, it fails to catch up with the experimental curve. The fast increase in the measured R H R when the fire is close to flashover is probably due to the downward flame spread, which was not reproduced well by the theoretical simulation. A very fine grid at the flame front could be helpful to predict the downward flame spread more accurately.

4.2 Gas temperature The gas temperature was measured at different locations and heights. Comparisons between the experiments and predictions are shown in Fig. 8. Generally there is good agreement, although some disparity exists. Figure 8 shows that the calculated temperature in the low gas layer and that close to the interface of two layers is significantly lower than that measured. This might be due to the radiation effect on the measurement thermocouple. Fortunately this will not affect the overall dynamics of the fire.

222

Z. }'an, G. Holmstedt

70

+

Point 1 exp

6O

......

pre - c

- - p r e

G 50

+ +

-p

+

+ +

+

70

......

~- 60

++

6_ 40 E 30

8O

Point 2 exp

- - p r e

+

+

-p

+ +

+

°~ 50 4+ +

: : : ,~,-I- +-

~- 30

20

20 10

10 t

I

I

100

200

300

0

I

I

I

100

200

300

Time (s)

250 200

+

Point 3 exp

......

pre - c

Time (s)

,

/

pre- p

O 6_ E

+

pre-c

" / '

150 100 ~.t.~+.'t~'

-

I 0

+

Point 5 exp

500

.....

pre - c pre-p

o 400

Point 4 exp

500

.....

pre - c o

~300 F--

200 100

I

i

100 200 Time (s)

600

+

G" o 400

./+'+~ . +++

50

600

0 0

300

t

I

200

300

Time (s)

+

80 + + . +++r~.'~; +

I

100

......

70

- -

~-60

.~

Point 29 exp pre - c

,,"

-pre - p

+ +

f /

°'-4 50

o

~' 3 0 0 ~- 200

•

I-- 30

100

10 I

I

100 200 Time (s)

Fig. 8.

I

300

0

i

0

-

100

I

q

200

300

Time (s)

Comparison of predictions and measurements of gas temperature. The symbol (c) denotes using Cone data input method and (p) the pyrolysis model.

223

Fire growth on wall lining materials

400

+

350

......

G 300 250 E 200 1-

. p,e-p/ 500 020030o

Point 34

Point 30 exp

700

pre - c

]

600

......

pre - c pre

. ~'

o~ +~47

150 100 - - , ~ - ~~+

50

I

0

600 d 400

E

t

{"

I

200

pre - c +

;'"

I

300

Point 51 ,-°

-I-

160

"

exp

/

...... ;'r::; /:'

¢i 120

300 t

I

100 200 Time (s)

300

Point 35 exp ++++

+ ......

200 F

I

1O0 200 Time (s)

700 T

1 O0

+

.+,+,,p,p~...i.,.1. "r •

=E loo

.~_,

~-

-

80

°°++

60 40 20

100 0

~ J 1O0 200 Time (s)

0

i 300

0 0

I

I

I

I

50

1O0 Time (s)

150

200

Point 54 600 T

+ 500~- . . . . . .

Point 55 600

exp pre-c

~" 400 -~

pre- p

~,+ + ,'1""~."

~" o 400

,."

Ec~ 300

~'300

-I......

500

P- 200

exp pre - c p,e-

-P.:'P,

~.+/~."

200

100I/~ ''F 0 0

J 50

100 I 100 Time (s)

I 150

~ 200

Fig.8.

0 0

(Continued.)

I

I

I

I

50

1O0 Time (s)

150

200

224

Z. )fan, G. Holrnstedt

4.3 Solid surface temperature The surface temperature of the material was measured on both the ceiling and the wall at different surface locations. About 30 comparisons are presented in Fig. 9. The theoretical data agree with the experiments fairly well. In the full-scale case, the calculated surface temperature is generally lower than the experimental data. This may, to a large extent, be the result of the underpredicted flame spread and fire growth at the later fire stages, as shown in Fig. 7(a).

4.4 Radiation flux In the duplicated experiment, the water-cooled radiation meter was used to measure the radiation flux at the floor, rear and left side walls, as shown in Fig. 6. The predicted radiation flux is compared with the results of these measurements in Fig. 10, in which reasonably good agreement can be seen. However, there is a considerable difference at the later stages of the fire at the measurement points 44 and 46. Due to the complexity of the problem, the reason for this is not very clear. This may mainly be because the sooting phenomenon was not considered adequately and the temperature field distribution was not reproduced very accurately. Soot makes its contribution to radiation through the radiation property. Since radiation intensity is proportional to the fourth power of temperature, the temperature profile can affect the radiation flux quite seriously.

4.5 Char depth (pyrolysis model only) When using the pyrolysis model to describe the burning of solid fuel, the mass release rate can be calculated. Therefore, for charring material, the char depth is available. In Fig. 11, a comparison of the measured and predicted char depth of the one-third scale fire is presented. In the experiment, the fire was extinguished about 4min after it started. The lining material was then removed and the char depth was measured by scraping the char layer from the material surface. In Fig. 11, the calculated char depth is represented by the colour and the measured value is indicated by the numbers. The comparison shows the predictions agree reasonably well with the experimental measurements. Both the calculations and the measurements indicate that the thickest char layer is located at the corner above the gas burner. Williamson e t al. 32 studied the char pattern of gypsum wallboard produced by a 5 min exposure at 40 kW of gas burner followed by a 1 min exposure at 150 kW in the full-scale room

Fire growth on wall lining materials

180 160 140

+

......

Point 6 exp

+..+

. +.'.//

pre - c pre - p

"P./

0 120

pre - c pre - p

s

. ~ 1 -

o

E

t

350

I

+

Point 8 exp

.....

pre- c

G 300 250

I

.,.+ +~,-..~

pre-p

~

"

I

I

Point 10 - . . . . pre - c

y

I

300

.4.,+ +,4r

e

0

B00 700 ~ 600

,,

pre - p

I

0

30O

.y

350 • 300 E 250 t9 ~- 200 150 100 5O 0

[

1O0 200 Time (s)

Point 9 500 + exp 45O . . . . . pre- c 400 + ~+" pp o 350 ~ 300

I

100 2O0 Time (s)

-

I

0

+

250 ~- 200 150 IO0 50

100 50

0

300

E ¢g 200 t- 150

,

°

I

1O0 200 Time (s)

+

I

300

Point 11 exp

. . . . . pre-p-reCl~..,~+"l"+'~+

o 500

E 400 (9

~- 300 2O0 100

I

0 9.

......

lID

400 -

Fig.

200

Point 7 exp

"~. 150

100 200 Time (s)

500 45O 4OO

+

~"

+'"

-100 80 I-60 40 20 0

250

225

I

100 200 Time (s)

I

300

0

0

I

I

I

100

200

300

Time (s)

Wall and ceiling surface temperature comparison. The symbol (c) denotes using Cone data input method and (p) the pyrolysis model.

226

Z. Yan, G. Holmstedt

Point 13

Point 12 600

600

=.="

+

+ + + J "

5OO °o

4O0

E CD 1--

300

+

~

500

g 400 d_

+

,

300

exp

+~

+ exp ...... pre-c -pre - p

I--

......

2O0

F

pre - c

- - p r e

200

-p

100

100 I

I

i

1 O0 200 Time (s)

100

300

300

Point 15

Point 14 700

700

+++++++

600

o

200 Time (s)

~

600

500

500

+ +,/" ~"

~- 400 E

300

~ 400 ~- 300

exp

÷ ......

2O0

pre

-

200

p

+

.+~,,~ +~"~'f..

pre - c

exp

- .....

',

pre - c -

pre- p

100

100 -

I

0

I

100

I

200

0

I

I

--I

1O0 200 Time (s)

300

Time (s)

300

Point 17

Point 16 7OO

700 600

++

~

÷÷÷y

__ 600500

G 500 o

400 E i~ 300

+ +

2OO

,r~/,,

d. 400 E i~. 300

exp

......

pre - c

"

200

pre - p

100

100 I

0

I

1O0 200 Time (s)

I

. y +~

y

0

300

Z

+

(Continued.)

pre - c pre - p

I

I

I

100

200

300

Time (s)

Fig. 9.

exp

......

Fire growth on wall fining materials

Point 18 800 700 600 0 o 500

Point 19 500 45O

? v

E 4OO ¢D 300 20O IO0

/

+, ~.p~

+ ......

pre - c

-

pm - p

-

I

+ ......

I

350

t

0

/

1O0 200 Time (s)

Point 20

300

Point 21 •

L) ° 400

+ . 1 "

0

300

600 + ......

exp pre - c pre - p

~. 300 E 250 2O0 150 100 5O

exp

100 200 Time (s)

500

227

exp pre-c

++

+ + 7' /

pre P

+

/

8O0 700

+

exp

I-v°~'o5E400ci 00500 "

E o 300 I--

2O0

++.+..+.#. ++~

100

I

0

I

t

100 200 Time (s)

0

300

0

+

Point 22 exp

/

700

500 ~- . . . . . .

pre- c

j,,

600

600 T

t.S

200 300 loo

~ 4 0 0 ~- ~ - p r e - p

.i.'t'/~, '

300

°v~"500

1O0 200 Time (s) Point 23 exp -I- "pre- c

300

. p ~

pre-p÷%~-" ++ ~Y

E 300

200~-

+

loo +

~.-t'¢, "

4

~

0 ~"+"+'~ J J 0 100 200 Time (s) "

200 100

j 300

~g. 9. (Continued.)

i

t

1O0 200 Time (s)

t

300

Z. Yan, G. Holmstedt

228

700

+

600

......

Point 24 exp pre-c

. + _+ +-~ -"

pro- p + +

......

5O0

,

4-+' ,~/,~'/

6" 500

+

600

~400

Point 25 exp pre-c pre - p

I=

+

300

+

200

-'"

~.,~ /

4"

6. 4OO E ~

++++'+

o*

/

.

~ 200 100

100 0

I

1

I

100 200 Time (s)

+ ......

500 450 400 O 350 6. 300 E 250 200 150 100 50 0

Point 26 exp pre - c

+ +

0

300

++

+

450 400 350

.",

......

d. 250 E 200

I--

+.'1" ~ " " I

I

I

1O0 200 Time (s)

-I-

Point 36 exp

200 L.~_. . . . . .

pre - c

250 7-

G ~ o_. Q. 150 ~-

pre - p

(

Point 27 exp pre - c pre -

300

+

+

o-

p + + ? ~

t/'" ,v?i .~."~"

150 100 50

I

0

0

300

I

100 200 Time (s)

I

I

100 200 Time (s)

300

Point 37

..;. .'..~" J "

350

+

300

......

pre - c - pre - p

G 250 o6. 200

100

exp .'F -At

_ . ~ /

.

/

~_ 150 100

0

t

0

r

100 200 Time (s)

~

0

300 F i g . 9.

0

(Continued.)

100 200 Time (s)

300

Fire growth on wall lining materials

700

+

600

......

~" o soo cL 400 E i~ 300

Point 39

Point 38 e~p

50O

pre - c ~ . . / ~ f" pre - p

rI-S

+

~ 450350 400

exp

......

pre - c pre - p

a. 300 E 250

o

20O

229

" + + , ~ .

÷:.

~ 200 150 IO0

+

100

I

0

100 200 Time (s)

8OO 7O0

.

+ . .

Point 40 exp . . .

300

1

I

1O0 200 Time (s)

300

Point 41 800 .

.

.-"

700

~600 0 50O

+

+ . ~

"~~ " " '

v 500 y

~ 40o

E 400 P- 300

~-a00

200

exp

+

......

200

100

pre - c pre

-

p

10(1 f

I

1

100 200 Time (s)

0

300

I

0

I

100 200 Time (s)

Point 42

I

300

Point 43

700

800

600 + + , + o,

G 500

~ 600

o v

o. 400 E 30o 20O

+/] ,'"

......

,,,

pre - c

/

~. 300 200

pre - p

100

IO0 I

I

~

100 200 Time (s)

0

300

Fig. 9.

.

......

pre - p

I

0

(Continued.)

pre - c

I

100 200 Time (s)

I

300

Z. Yan, G. Holmstedt

230

Point 57

Point 56 700

900 800 700 600 500 E (1) 400 I300 200 100 0

+ ~ - I - . o-.

.A ~ / r

+

+

~-. 600500 ~.

° o o

exp

......

+ + +

pre - c pre - p

f

400

ex.

. . . .

pre - c

~

pre - p

200

100 0

I

t

t

50

100 Time (s)

150

--~1 200

--

0

t

I

P

I

50

10O Time (s)

150

200

Point 61

Point 59 600

4-

+ ......

500

exp + pre - c + p or ~ . . ~

P 400 v

E 300 I1) I-200

I.i./~-.,. .,+.,,+1)"

100

0

500

~ o ,-; E ~-

•

+

450 400 350 300 250 200

exp

......

+

pre - c pre - p

+

+ / j , , ~"

.e a

15o

100 50

J

I

I

t

-

P

I

I

I

50

10O Time (s)

150

200

0

50

100 Time (s)

150

200

Fig. 9.

(Continued.)

corner fire. On the side walls, they found there is substantial charring up to 1.5 m in height as well as some charring at the ceiling level. This kind of pattern was qualitatively reproduced well by the CFD simulation which shows that at a certain time, on the side walls above the gas burner, the char depth and surface temperature at the ceiling level are higher than those at the middle level. This is also indicated by our experimental temperature measurements.

4.6 Mass flux through opening In the full-scale experiment 3~ the mass flux through the doorway was measured with three bidirectional probes located above the neutral plane. The position of each probe was chosen based on experience from other experiments, in order to achieve as good an estimate of the total mass flux as possible. The measured results, however, are crude due to the few

Fire growth on wall lining materials

Point 45

Point 44 70 6O X = ~ 50 : . ~ 40 ~ ~.~- 30 n20

exp

+ ......

25

/

pre - C ~

°

/

++ + +

x

20

=~"

15

+ ......

t t

oE10 rr

+ ......

.o-/

..4X,

5

, 300

•

+

0 100 200 Time (s),

300

Point 4 7 +

exp pre-c

t

i

pre - c -pre-p

Point 46 14

exp

~ ~ 10 rr

10 " T+ , 0 / :+~+.~ 0 1O0 200 Time (s)

16

231

+ ." +.,J

~

p r e y . .

18 I + 16 | 14~ . . . . . .

~'~" 12+10

exp (by B. Andersson) pre-c pre-p

÷ .:-I-/

.~

6 4

0

+

0

100 200 Time (s)

300

.

exp

......

pre- c

~pre-p

.l+j/ , ,~;/

6

0

I

0 Fig. 10.

300

Point 48

18 16 X ..~ 14 E ~ 12

rr

100 200 Time (s)

t

100 200 Time (s)

I

300

Predicted and measured radiation flux. The symbol (c) denotes using Cone data input method and (p) the pyrolysis model.

Z. Yan, G. Holmstedt

232

,3.871 mm 2.904 1.936 0.968 0.000

(a)

\

j

~

\

113.871

(b)

mm

m2.9o4 1.956 !::: 0.968 n 000

~

I

\ Fig. 11.

Predicted and measured char depth of one-third scale fire: (a) char depth on the walls; (b) char depth on the ceiling.

233

Fire growth on wall lining materials

(a)

(b) 0.87

400 exp pre(c)

07 ~ 0.64

- - - pre(p)

,-/''--/ /

I

/ --/

~, 350 ~

j Y

0.5 ~ O e-

2so 2oo

0.4.o

0.3: 0.2-

300

/"

/,

exp -pre(c) - - -pre(p)

,'/ p

150

~= 1001

0,1

>.

50: 60

120

180

T i m e (s)

60

120

180

T i m e (s)

Fig. 12. Mass flux and convection heat loss through opening. The symbol (c) denotes using cone data input method and (p) the pyrolysis model. measuring points. Figure 12(a) presents the comparison of calculated and measured mass flux through the doorway. The measured data are consistently higher than the prediction, and the curves are almost parallel to each other. One important reason for the difference between the experimental measurements and the predictions might be the crudeness of the measurement. The good agreement of the convection heat loss through the opening suggests that the possibly underpredicted entrainment could be another reason.

4.7 Convection heat loss through opening In the full-scale experiment, the convection heat flux through the door was estimated together with the mass flux. The calculated data are compared in Fig. 12(b) with the experimental data. Although there is good agreement, this comparison can only be taken as a rough guide, due to the uncertainty involved in the experimental estimate.

4.8 Heat release rate per unit area and graphical presentation 4.8.1 H e a t release rate p e r unit area As mentioned above, due to the transient incident heat flux, the burning behaviour would be quite different in a real fire scenario from that in the Cone test. Figure 13 shows the heat release rate per unit area, calculated using the pyrolysis model, of the solid elements close to the gas burner. Due to the higher heat flux, the R H R per unit area in the full-scale fire is consequently higher than that in the one-third scale fire. Comparing this figure with Fig. 1 shows considerable difference between them.

Z. Yan, G. Holmstedt

234 (a)

(b) 180 160

120

140 100

120 ~"

100 v ,.¢

80

~ 6o

8060

"1" e~

40

40 20

20 60

Fig. 13.

120

180

0

60

120

180

240

300

Time (s) Time (s) Typical RHR per unit area calculated with pyrolysis model in the simulation of fire growth: (a) full scale; (b) one-third scale.

In Table 1, the average heat release rate per unit area is given. The average heat release rate per unit area is obtained by dividing the total heat release rate by the burning area. Due to the difficulty of determining burning area during a fire, only one set of burning area data is available for each test, which can be measured from the burnt material after the experiment. From this table, it can be seen that the average R H R per unit area in the Cone data input method is much higher than both the experimental data and the result obtained from the pyrolysis model. The result from the pyrolysis model agrees reasonably well with experiment. This implies that the promising prediction of total RHR obtained with the Cone data input method seems to be attributable to the compensation of the small burning area to the high Cone Calorimeter RHR input. The pyrolysis model is able to give a better description of the response of the solid material. In the experiments, photographs were taken and video recordings were made. The area involved in the flaming combustion was estimated afterwards from the video recording or photograph. However, due to TABLE 1 Average heat release rates per unit area

Scenario

Full scale One-third scale

t = 180 s t = 260 s

Experiment

Cone method

Pyrolysis model

No data 60 k W / m z

176 k W / m 2 139 k W / m 2

98 k W / m 2 80 k W / m 2

Fire growth on wall lining materials

235

smoke obscuration, the accuracy of this estimated area is questionable. Dividing the measured total R H R by this estimated area gives even lower values of the average RHR per unit area--only about 40 kW/m 2.

4.8.2 Graphical presentation Figure 14 shows the calculated wall and ceiling surface temperatures of the one-third scale fire. From this figure, the pattern of fire spread along 1200 K q75

Fig.

14.

Surface temperature distribution.

236

Z. Yah, G. Holmstedt

the wall behind the burner, the ceiling-waU intersection and on the ceiling can clearly be seen. When using the pyrolysis model, the long side wall opposite the gas burner was ignited (this cannot be seen in the figure) at about 220s after the fire started. This agrees reasonably well with Andersson's 3~ experiment where the downward flame spread on this wall was observed at a time of 255 s after the ignition of the gas burner.

5 CONCLUSIONS The flame spread over surface lining material has been simulated using the CFD method. The simulations generally compare favourably with experiment. The heat release rate, specimen surface temperature and gas temperature are satisfactorily reproduced. However, there are some differences between simulation and experiment regarding details. This implies that further research is required and the improvements in the following areas are expected. The physical properties of solid materials, such as specific heat and conductivity, have significant influence on the flame spread. These properties usually change with temperature. To provide accurate values of these properties as an input for CFD simulation, more thermo-physical tests on solid combustible materials are required. It would be advantageous to construct a data base containing such information. It was shown that the total heat release rate and the temperature environment can be reasonably well reproduced if the Cone Calorimeter input data are chosen well. However, this seems to be attributed to the compensation made by the small burning area to the high Cone Calorimeter R H R input. As demonstrated by this preliminary study, a transient pyrolysis model giving a better and more reasonable description of the response of solid material would be desirable. The burning behaviour of solid materials changes with the incident heat flux and might be quite different in a real fire scenario from that in the Cone Calorimeter test. By using a pyrolysis model, the important effect of the transient heat flux on the burning of solid material can be considered. Different materials have quite different sooting behaviour. Consideration of this should be included to improve the prediction of soot mass concentration which significantly affects the prediction of radiation. Perhaps the most important conclusion is that the studies carried out indicate that the flame spread in a compartment fire from ignition to flashover can be expected to be simulated by CFD.

Fire growth on wall lining materials

237

ACKNOWLEDGEMENT This work was supported by the Swedish Fire Research Board ( B R A N D F O R S K ) , which is gratefully acknowledged. We also wish to express our special thanks to Professor G. Cox, Dr S. Kumar (Fire Research Station, UK) and John deRis (FMRC) for valuable discussions.

REFERENCES 1. Frey, A. E. & T'ien, J. S., A theory of flame spread over a solid fuel including finite rate chemical kinetics. Combust. Flame, 36 (1979) 263-287. 2. Di Blasi, C., Crescitelli, S., Russo, G. & Fernandez-pello, A. C., Prediction of the dependence on the opposed flow characteristics of flame spread rate over thick solid fuel. In Fire Safety Science--Proc. of the 2nd Int. Symp, 1988, pp. 119-28. 3. deRis, J. N., Spread of a laminar diffusion flame. In 12th Syrup. on Combustion, Combustion Institute, 1969, pp. 241-52. 4. Quintiere, J. G., A simplified theory for generalizing results from a radiant panel rate of flame spread apparatus. Fire Mater., 5 (1981) 9. 5. WickstriSm, U. & G~ransson, U., Prediction of heat release rates of surface materials in large-scale fire tests based on cone calorimeter results. J. Test. Eval., 156 (1987) 364-370. 6. Cleary, T. G. & Quintiere, J. G., A framework for utilizing fire property tests. In Fire Safety Science--Proc. 3rd Int. Symp., 1991, pp. 647-56. 7. Karlsson, B., Modeling fire growth on combustible lining materials in enclosures. Ph.D. Thesis, Department of Fire Safety Engineering, Lund University, Sweden, 1992. 8. Opstad, K., Modelling of thermal flame spread on solid surfaces in large scale fires. MTF-Report 1995:114 (D), Department of Applied Mechanics, The University of Trondheim, 1995. 9. Opstad, K., Modelling of thermal flame spread on solid surfaces in large scale fires. In First European Symp. on Fire Safety Science, 1995, pp. 93-4. 10. Yan, Z. H. & Holmstedt, G., CFD simulation of flame spread in room fire. In First European Symp. on Fire Safety Science, 1995, pp. 95-6. 11. Cox, G. & Kumar, S., Field modeling of fire in forced ventilated enclosures. Combust. Technol., 52 (1987) 7-23. 12. Fan, W. C., Yan, Z. H. & Jiang, X., Numerical simulation of fire process in zerogravity environment. In Proc. 2nd Asian-Pacific Int. Symp. on Combustion and Energy Utilization, 1993. 13. Kumar, S. & Cox, G., Mathematical modeling of fires in road tunnels. In 5th Int. Symp. on the Aerodynamics and Ventilation of Vehicle Tunnels, 1985, pp. 61-76. 14. Rodi, W., Turbulence models and their application in hydraulics--a state of the art review. SBF Report 80/T/125, University of Karlsruhe, 1980. 15. Magnussen, B. F. & Hjertager, B. H., On mathematical modeling of turbulent combustion with special emphasis on soot formation and combustion. In 16th Int. Syrup. on Combustion, The Combustion Institute, 1976, pp. 719-29.

238

Z. Yan, G. Holmstedt

16. Lockwood, F. C. & Shah, N. G., A new radiation solution method for incorporation in general combustion prediction procedures. In 18th Int. Symp. on Combustion, The Combustion Institute, 1981, pp. 1405-14. 17. Modak, A. T., Radiation from products of combustion. Fire Res., 1 (1978) 339-361. 18. Hottel, H. C. & Mangelsdorf, H. G., Trans. Am. Inst. Chem. Engng, 31 (1935) 517. 19. Hottel, H. C. & Egbert, R. B., Trans. Am. Inst.. Chem. Engng, 38 (1942) 531. 20. Leckner, B., Combust. Flame, 19 (1972) 33. 21. Mulholland, G. W., Smoke production and properties. In The SFPE Handbook of Fire Protection Engineering, 2nd Edn, Ch. 2-15, 1995. 22. Tewarson, A., Generation of heat and chemical compounds in fires. In The SFPE Handbook of Fire Protection Engineering, 2nd Edn, Ch. 3-4, 1995. 23. Moss, J. B., Modelling soot formation for turbulent flame prediction. In Soot Formation in Combustion (Mechanisms and Models), Vol. 59. Springer Series in Chemical Physicals, 1994, Berlin. 24. Tsantaridis, L. & Ostman, B., Smoke, gas and heat release data for building products in the cone calorimeter. Report I 8903013, Trateknik Centrum, Sweden, 1989. 25. Fredlund, B., A model for heat and mass transfer in timber structures during fire. Ph.D. Thesis, Department of Fire Safety Engineering, Lund University, Sweden, 1988. 26. Atreya, A., Pyrolysis, ignition and fire spread on horizontal surfaces of wood. NBS-GCR-83-449, National Bureau of Standards, 1984. 27. Sibulkin, M., Heat of gasification for pyrolysis of charring materials. In Fire Safety Science--Proc. 1st Int. Symp., 1985, pp. 391-400. 28. Quintiere, J. G. & Rhodes, B., Fire growth models for materials. NIST-GCR94-647, National Institute of Standards and Technology, 1994. 29. Delichatsios, M. A. & C h e n , Y., Flame spread on charring materials: numerical prediction and critical conditions. In Fire Safety Science--Proc. 4th Int. Symp., 1994, pp. 457-68. 30. Sundstr/Sm, B., Full-scale fire testing of surface materials. Technical Report of Swedish National Testing Institute, SP-RAPP, 1986, p. 45. 31. Andersson, B., Model scale compartment fire tests with wall lining materials. Report LUTVDG/(TVBB-3041), Fire Safety Engineering Department, Lund University, Sweden, 1988. 32. Williamson, R. B., Revenaugh, A. & Mowrer, F. W., Ignition source in room fire tests and some implication for flame spread evaluation. In Fire Safety Science--Proc. 3rd Int. Symp., 1991, pp. 657-66.