Flame Radiation in Large Fires Wighus R.*, Brandt A. W., Sesseng C. SP Fire Research AS, Fire Development and Mitigation, Trondheim, Norway *Corresponding author email:
[email protected] ABSTRACT A paradox of the assessment of the emissivity of the flames is that this is based on the measured flame temperature. The emissivity of flames is not a generic property that can be assessed without measuring the flame temperature, so the correct assessment of flame temperature is part of the chain to obtain the correct value. In many of the experiments reported in the literature, flame temperatures have been measured with thermocouples embedded within the flame envelope, assuming that the flame surrounding is close to optical thick, which is assuming a flame emissivity close to unity. This leads to a possible error of the assessment of flame emissivity, which consequently leads to uncertain assessment of radiation from flames.
KEYWORDS: Emissivity, large fires, radiation. NOMENCLATURE F km L T
ε σ
View factor (-) Specific emissivity factor (m−1) Flame dimension (m) heat flux density (kW/m2) temperature (K)
emissivity (-) Stefan-Boltzmann constant (kW/(m2·K−4)) Subscripts f flame r radiation
Greek BACKGROUND
In hazard evaluations, the impact of a fire is often quantified by a heat flux density towards an object (kW/m2), and this heat flux density is used as an input to calculations of heat load to external objects as well as objects embedded by the flames and the fire plume. It is also used to calculate the evaporation rate of liquid pool fires, and is an important factor in assessment of fire dynamics. An extensive summary of research in the area of thermal radiation from large fires is contained in the chapter in the Society of Fire Protection Engineers (SFPE) Handbook, entitled Fire Hazard Calculations for Large Open Hydrocarbon Fires [1]. In this summary, a statement about thermal radiation which is commonly used in the literature is given: For fires greater than a few meters in diameter, the effective emissivity of the flame can be taken as one. In other references like the NIST Report 6546 [2], this statement is: For fires greater than a few meters in diameter, the effective emissivity of the flame can be taken as one. Also, to be on the conservative side, the transmissivity is taken as one. A study of the flame radiation properties of open hydrocarbon pool fires was presented by Ufuah and Baily in 2011 [3], which sums up most of the published background for hazard assessment from such fire to the ambient. In Fig. 1, their model for radiative energy flux as a function of pool fire diameter is shown, together with three other models. The model for heat flux density that is presented Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8), pp. 296-305 Edited by Chao J., Liu N. A., Molkov V., Sunderland P., Tamanini F. and Torero J. Published by USTC Press ISBN:978-7-312-04104-4 DOI:10.20285/c.sklfs.8thISFEH.030
296
Part II Fire
is based on published data from experiments, and the method of least squares is used to fit the data. Shokri and Beyler [4] have a model that reduces the radiative energy flux with pool diameter after an initial phase with increasing flux. Mudan and Croce [5] show initially increasing flux with diameter, then a quasi-steady phase, and subsequently a proportional increase with diameter. McGrattan et al. [2] predict a steady radiative heat flux with increasing diameter. Ufuah and Baily have a model (denoted “Present model” in Fig. 1) with increasing radiative heat flux up to a pool diameter about 200 m, then a decreasing output. All these models predict the radiative energy flux (kW/m), which is the Surface Emissive Power of the fire plume (kW/m2), multiplied with the plume height (m). The different models are all based upon published experimental data, and it seems that the physics behind the models is insufficient to describe the complex phenomena of an open pool fire.
Figure 1. Radiative energy flux as a function of pool fire diameter, based on four different models [2-5].
Table 1 and Fig. 1 exhibit some of the experimental data that the model of Ufuah and Baily is based upon, and one can see that there is limited basis for validation of models for pool diameters above 20 m. The values in Table 1 also show non-consistent relation for heat flux density with varying pool diameter. We will not claim that any of the presented models are right or wrong, but will use this as an example of non-consistency of present models for radiative properties of large pool fires. This is also underpinned by the tables for Maximum Time-Averaged Temperatures and Heat Flux Densities, given in Table 3-11.8 and Table 3-11.9 in SFPE Handbook [6]. Maximum Time-Averaged Temperatures of Pool Fires varies from 770 to 1200 °C. Heat Flux Densities of Pool Fires spans from 75 to 170 kW/m2. The theoretical basis for treating heat load to remote objects, using both empirical models and complex geometrical calculations for radiative emission from flames, is not the topic of this paper, since we will concentrate on heat load to objects embedded by flames. The use of flame emissivity in assessing the heat flux density varies from different authors and textbooks. One of the basic textbooks, “Drysdale: An Introduction to Fire Dynamics” [7], touches the question of flame emissivity by mentioning that: A few empirical values of km is available in the literature, and permit approximate values of emissive power to be calculated, provided the flame temperature is known or can be measured. This presumption that the flame temperature can be estimated is focused and discussed in this paper.
297
Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8) Table 1. Heat flux density and pool fire diameter, used as a background for modelling radiative energy flux [3]. Q (kW/m2)
D (m) Gasoline
1.5
59.23 Munos et al.
Gasoline
3
76.24 Munos et al.
Gasoline
4
81.76 Munos et al.
Diesel
3
67.57 Munos et al.
Diesel
4
71.71 Munos et al.
JP-4
1.22
69.06 Fu
JP-4
2.44
44.63 Fu
JP-5
2.44
46.47 Fu/Dyan & Thien
JP-5
3.05
79.99 Fu/Dyan & Thien
JP-5
5.5
53.31 Fu/Dyan & Thien
LNG
14.64
49.06 May & McQueen
LNG
16.17
50.56 May & McQueen
LNG
18.3
53.76 May & McQueen
From the basic work of Griffiths and Awbery [8], we site: […] it would seem that a working definition, which corresponds to the general idea of what is meant by “the temperature of a flame”, would be “the temperature of a solid body which is in thermal equilibrium with the flame”. In their work, however, the methods of flame temperature measurement with an electrically heated wire as well as an optical method of measuring the flame temperature were used. Attempts were made to measure real flame temperatures at stoichiometric mixed reactants. The focus of this work was to obtain non-intrusive temperature measurements, and thermocouples were seen as possible intrusive objects, and that the inserted objects cooled the flame. From an operational point of view, there are as many different “temperatures” in flames as there are measuring techniques. SPECIFIC FLAME EMISSIVITY A variation of values given for the constant km, the specific flame emissivity, is seen in different textbooks and literature. Values as high as 5.3 for Polystyrene, 1.8 for Polypropylene and 1.3 for PMMA, are presented by Drysdale [9]. For liquid pool fires, Mudan and Croce [5] present values of 0.4 for LNG and 0.335 for LPG. One can see in Fig. 2 that using estimates of km with high values, like the ones that is obtained in laboratory experiments burning plastic fuels, flames will obtain high emissivity at moderate thickness. In large fires, like gas or oil leakages, the emissivity of a flame with thickness in the order of 1 m is only a fraction of this. This is in contrast to the commonly referred values, which is that a fire is known to be optically thick and turbulent, for pool diameters larger than 1 m [5, 7].
298
Part II Fire
Figure 2. Emissivity of flames as a function of effective emission coefficient km.
The relatively low values of Surface Emissive Power that is measured in contrast to the internal heat fluxes inside flames is speculated to have its origin in the blockage of radiation by sooty parts of the flames. The main issue is that the heat load inside flames varies with size of the fire, irrespective of the type of fuel and other fire characteristics. RESULTS FROM LARGE-SCALE EXPERIMENTS Since the beginning of the 1990s, the oil industry has been concerned about heat loads from oil- and gas leakages, and has carried out several large-scale experiments to obtain more reliable data for heat loads and erosive effects of this type of fires. Some of these data have been presented in industryrelated seminars and conferences, but a large amount of data is still not published. In the 3rd Edition of the SFPE Handbook, Beyler touches this when referring to high heat fluxes reported by Cowley and Prichard [10]: Heat fluxes of up to 250 and 300 kW/m2 for two-phase LPG and sonic natural gas jets, respectively, have been measured in large-scale jet flame tests. Wighus and Drangsholt [11] report temperatures as high as 1200 °C and impingement velocities of up to 80 m/s in gaseous propane jet flames. They also found that the temperatures observed at the location of peak velocity were lower for higher gas velocities. They measured heat fluxes as high as 340 kW/m2 in some tests, and the radiative fraction of the total heat flux tended to be about 2/3. In 1994, a Joint Industry Project, Blast and Fire Engineering for Topsides Structures, was carried out, with partners from oil- and gas companies operating in the North Sea. Large scale tests were carried out both in UK and Norway, and interesting results were found in the Test Programme F3, Confined Jet and Pool fires. This was carried out by SP Fire Research (previously SINTEF NBL), in cooperation with Shell Research, Thornton UK, [12, 13]. The dimensions of the test compartment were in most cases approximately 6 m high, 6 m wide and 12 m long (net. 415 m3), with an opening of variable size at one short end. The fuels were in most cases condensate of crude oil from the North Sea, and vertical or horizontal spray fires with a release rate of 0.9 kg/s were tested. In some cases, liquid pool fires of 24 m2 size were tested. The average maximum temperatures observed in the tests were 1258 ± 80 °C (N = 19). Maximum temperature observed in the tests were 1370 °C, which is the maximum measuring range for the thermocouples used, indicating even higher temperatures. 299
Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8)
Heat flux densities measured in the 415 m3 test compartment are shown in Table 2. Table 2. Enclosed fires – Results from experiments [12]. Fire type
Heat load
Jet fire hitting an object
Local heat loads of 350-400 kW/m2. Typical heat loads to walls 200 kW/m2
Enclosed pool fire
A ventilation ratio of 0.8 of stoichiometric mixture gives the highest evaporation rate. Local heat load values of 300-350 kW/m2 were measured in a room volume of 415 m3 and a 24 m2 condensate pool area. Typical heat load to walls 200 kW/m2
OPEN LIQUID POOL FIRES In 1996 SP Fire Research carried out a series of crude oil pool fires on the sea surface, to study the fire characteristics and the heat load onto structures engulfed in such fires [14]. These tests were carried out at Spitsbergen, Svalbard, Norway. The first test series (the winter tests) was carried out in a 15 m diameter basin cut in the 1 m thick ice, to obtain wind conditions as on the sea, as seen in Fig. 3. A 1 m diameter steel target, reaching about 8 m above the sea, was instrumented with a large number of thermocouples to measure the steel temperature. Heat flux meters were also inserted in the target surface, and outriggers with additional thermocouples were attached to the target. Gas concentrations (O2, CO2 and CO) were measured from suction points inside the flames.
Figure 3. The crude oil fire of the winter tests, 8000 litres burning in a 15 m diameter basin [14].
Thermocouple temperatures increased rapidly and reached above 1200-1300 °C within less than 5 minutes. The measured temperatures exhibit similar features as was seen in the enclosed condensate spray and pool fires, with at least one of the thermocouples reaching 1370 °C, the system saturation temperature. Other temperatures were well above 1200 °C. A fireball of 4-6 m diameter appeared in the zone where the highest temperatures were measured, and the burning seemed to exhibit recirculation and intense mixing. The main results are given in Table 3.
300
Part II Fire Table 3. Open fires – Results from experiments [14]. Fire type
Heat load
Pool fire on the sea surface –8000 In the lower 1/3 of the flame, a 30 seconds time average heat flux litres of crude oil, 15 m diameter was measured 260 kW/m2. The 60 seconds average was 220 kW/m2 pool
From all these large-scale test, it seems more likely that flame emissivity is increasing with diameter and obtains a value closer to unity first at diameters more than 6-10 m. This indicates that the values of effective emission coefficient km will be in the range 0.3-0.4. It is possible that these values are even lower, and further studies are needed to investigate this. However, measurement of radiation inside flames is complex and challenging, and there is a need for developing robust and reliable measurement techniques. STANDARD FIRE LOAD Many of the predictive models in fire science are so-called empirical models, based on observations and measurements in experiments and in a few cases, in accidental fires. The consequences of a fire are to be taken into account in risk analyses, in the dimensioning of firefighting measures and fire protection engineering. In many cases, standardized fire tests are developed and adapted into the performance requirements for buildings and constructions. Among the fire test standards, the socalled ISO 834-curve for fire resistance tests is wide-spread and recognized worldwide [15]. This ISOcurve describes a time-temperature history to be reproduced in test furnaces. This is to ensure that products and construction elements will be subjected to a reproducible test wherever it is carried out, and independent of the operators and laboratory.
Figure 4. Temperatures measured by thermocouples installed at different distances from the fire source, in the fire plume inside the Runehamar test tunnel, [16].
However, one side-effect of the agreement of this time-temperature curve for fire resistance testing is that the curve is used as a universal representation of fire temperatures in the real world. The ISO 834 Cellulosic fire curve was in the 1990s supplemented by the Hydrocarbon fire curve with a maximum temperature of 1100 °C representing the fire severity of the oil- and gas industry [17]. In the 301
Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8)
Netherlands, an even more severe time-temperature curve with a maximum temperature of 1350 °C, used for testing tunnel linings, has been launched. This curve, is based on tunnel fire tests carried out in the Runehamar test tunnel in Norway [16]. The tests were conducted in a tunnel with a standard profile, with approximately 50 m2 cross section area. The fuel was a setup with up to 720 pallets organized like a truck load. Most of the pallets were made from wood, a small amount were plastic. Again, the fire in an enclosure of a certain size, leads to temperatures well above 1200 °C. An example of thermocouple temperatures from the Runehamar tunnel experiments is presented in Fig. 4. FLAME TEMPERATURE AND RADIATION The “flame temperature” used in hazard assessment varies a lot in literature, and the two most common definitions can be formulated as 1) The flame temperature is interpreted as the temperature an object embedded by a flame obtains at steady-state. [8], and 2) The flame temperature is interpreted as the temperature which represents the Surface Emissive Power of a flame (measured remotely), assuming a known shape of the flame, the actual view factor and an assumed flame emissivity. HEAT TRANSFER TO OBJECTS ENGULFED BY A FIRE A fire, like a hydrocarbon liquid pool fire or a fire-ball from a gas release, is a very complex and dynamic process influenced by chemical and physical parameters. The mixing of reactants and the resulting stoichiometry of the combustion process is a challenge to those who want to model the processes mathematically, even when complete mixing is assumed. In large fires, different parts of the flames will have different mixtures of reactants, and even if the combustion processes takes place in fractions of seconds, some of the chemical reactions have time scales which may enhance or reduce the completeness of combustion inside the flame envelope. This will consequently lead to temperature variation over the flame volume. The entrainment of surrounding atmosphere into the fire plume varies with the height above ground for fire plumes in non-wind conditions, and wind influences the temperature field and the concentration field of combustion reactants and products. The simplified model for heat load from a fire plume or a fire-ball to an object assumes that the flame is to be seen as a solid body. The size and geometry of the body is normally assessed by transforming the flame into a symmetrical object, either as a cylinder, a cone or a sphere. The border between the object and the surrounding atmosphere is often taken as the extent of the visible flame, deduced by photography. The simple model then treats the flame as a solid, with a surface described by the geometry model, and an average temperature. The heat flux density from this surface is then measured by heat flux metering techniques, either calorimetric devises or devises measuring radiation (mostly infrared radiation measurements). Fig. 5 illustrates a set-up where surface emissive power of a pool fire is interacting with an object. Fig. 5 shows (in 2D) a thermocouple, located inside a flame, receiving and emitting radiation with its surroundings. Section denoted I illustrates that the thermocouple exchange radiation partly with the flame and partly with the base of the fire, section II will exchange radiation with the flame and the ground. Section III will exchange radiation with the flame and the radiation receiving/emitting object, whereas section IV will exchange radiation with the flame and the atmosphere. The radiation exchange with the objects and surroundings is dependent on the optical density of the flames. If the flames are considered to be optically thick (emissivity of 1.0), no radiation exchange will occur between the thermocouple and objects outside the flames. If the emissivity is 0.5, the exchange with the surroundings may count for 50% of the exchange. The flame in Fig. 5 is divided into zones, illustrated by the circles. One may consider these zones as objects, with a certain temperature and emissivity giving a fraction of radiation and re-radiation. If
302
Part II Fire
the flames become sufficiently thick, the surroundings will lose the influence on the thermocouple temperature.
Figure 5. Illustration of a thermocouple, located inside a flame, receiving and emitting radiation with its surroundings.
RADIATION TRAPPING AND ENCLOSURE EFFECT In an enclosure with well insulated walls, the surface temperature of the surroundings may become high, and the radiation exchange with the flames will be reflecting real flame temperatures. Similar occurrence will be obtained in optically thick flames, since the radiation exchange between the thermocouple and the flames will not be influenced by the surroundings. This is in some cases denoted radiation trapping, described by Chamberlain, Persaud, Wighus and Drangsholt [12, 13]. The maximum temperature of complete combustion is the adiabatic flame temperature, which will be obtained when the combustion can take place with exactly stoichiometric conditions and without any loss of energy from the flames. To obtain this temperature, a test apparatus with insulated walls is constructed. When the wall temperature is similar to the adiabatic flame temperature and no energy transport takes place through the walls, all the chemical energy produced by combustion is transferred to the combustion products, and leads to a stable gas temperature. When a thermocouple is mounted inside such an apparatus, a steady-state condition of the thermocouple is obtained. No convective heat transfer takes place since the surface temperature of the thermocouple is similar to the gas temperature. Radiation to the thermocouple balances radiation from the thermocouple to its surroundings. In fire test furnaces, the only heat loss from the combustion is by the combustion products (smoke from the furnace) and heat transported into the test object. Radiation exchange between a thermocouple inside such a furnace and in theory the flames should ideally lead to close-to-adiabatic temperature. In practice, the fire test furnace temperature is regulated by the air-and-fuel supply. This mechanism makes it is possible to follow different time-temperature curves. An open flame can be considered to act in a similar way as a furnace. With optically thick flames, the emissivity should approach unity, implementing that radiation from a thermocouple to the flames would balance the radiation from the flames. An equilibrium temperature balancing the ingoing and outgoing flux of energy should be obtained. However, if the flames are not optically thick, a fraction 303
Proceedings of the Eighth International Seminar on Fire and Explosion Hazards (ISFEH8)
of the heat flux from the flames will radiate to the surroundings. The surroundings may be the atmosphere, the ground below the flames or objects and constructions in the vicinity of the flames. It may also be parts of the flames with lower temperature and possibly sooty flames. This will lead to an equilibrium temperature lower than the actual flame temperature. FLAME EMISSIVITY – THEORETICAL MODEL The radiative heat flux is described by Stefan-Boltzmann law of blackbody radiation, as described in Eq. (1): (1) q& r¢¢ = F × e × s × T 4 , where F = the geometrical shape factor between the objects exchanging radiation, ε = surface emissivity of the radiating object, σ = Stefan-Boltzmann constant, 5.67 ´ 10−11 (kW/(m2·K−4)) and T = object temperature (K). For non-solid objects, like a fire plume, the Stefan-Boltzmann law has been modified by introducing a flame emissivity, described by Eq. (2):
e f = 1- e-kmL ,
(2) −1
where km = effective emissivity constant, estimated empirically (m ) and L = the characteristic dimension of the radiating object (the grey body), e.g. the flame thickness (m). DEPENDENCY OF MEASURED FLAME TEMPERATURE ON EMISSIVITY The estimation of flame emissivity is in most referred situations based on thermocouple temperatures, and the thermocouples are mounted inside the flames. When the emissivity is estimated, one normally will use Eq. (1) combined with Eq. (2), and we can find an equation for the specific emissivity constant k m: qr ¢¢ æ ln ç 1 s × F ×T 4 è km = L
ö ÷ ø.
(3)
In this equation one can see that the estimate of specific emissivity constant is based on the measured flame temperature and the dimension of the flame. If flame temperature is measured by an embedded thermocouple or any similar type of object that is exchanging radiation with the flames and its surroundings, the measured temperature itself is a function of the flame emissivity. This means that the assessment of effective emission coefficient is dependent on two factors that are not measured independently. This means that there are a variety of pairs of flame temperatures and length scales that can fulfil the equation and a real independent effective emission coefficient cannot be established by this type of measurements. CONCLUSION A severe lack of information on radiative properties of flames is pointed out in this paper. The statement of Drysdale: “A few empirical values of km is available in the literature, and permit approximate values of emissive power to be calculated, provided the flame temperature is known or can be measured” still leaves an open question, since most of the measured flame temperatures in fires are biased by their surroundings, and based on an unrealistic assessment of flame emissivity. The measurement of thermocouple temperatures well above 1300 °C in fires of different types, like open and enclosed hydrocarbon pool fires, spray and jet fires and even cellulosic fires should lead to a revision of how fire size influences radiative properties of fires. The assessment of specific emissivity constant for fires should not be based on flame temperature measurements with probes that are dependent on the emissivity.
304
Part II Fire
REFERENCES 1. Mudan, K. S., and Croce, P. A. Fire Hazard Calculations for Large Open Hydrocarbon Fires, SFPE Handbook of Fire Protection Engineering (2nd ed.), National Fire Protection Association, 1995. 2. McGrattan, K. B., Baum, H. R., and Hamins, A. P. Thermal Radiation from Large Pool Fires National Institute of Standards and Technology, Gaithersburg, USA, NISTIR 6546, 2000. 3. Ufuahe, E. and Baily, C. G. Flame Radiation Characteristics of Open Hydrocarbon Pool Fires, Presented at the World Congress on Engineering, London, UK, 2011. 4. Shokri, M., and Beyler, C. L. Radiation from Large Pool Fires, Journal of Fire Protection Engineering, 1(4): 141-150, 1989. 5. Mudan, K. S., and Croce, P. A. Fire Hazard Calculations for Large Open Hydrocarbon Fires, SFPE Handbook of Fire Protection Engineering (1st ed.), National Fire Protection Association, 1988. 6. Di Nenno, P. J. SFPE handbook of fire protection engineering (1st ed.) National Fire Protection Association Quincy, MA, 1988. 7. Drysdale, D. An Introduction to Fire Dynamics, Wiley & Sons, 1999. 8. Griffiths, E., and Awbery, J. H. The Measurement of Flame Temperatures, Proceedings of the Royal Society London, 123(792): 401-421, 1929. 9. Drysdale, D. An Introduction to Fire Dynamics, John Wiley and Sons, 1985. 10. Beyler, C. L. Fire Hazard Calculations for Large Open Hydrocarbon Fires, SPFE Handbook of Fire Protection Engineering (3rd ed.), 2002. 11. Wighus, R., and Drangsholt, G. Impinging Jet Fire Experiments-Propane 14 MW Laboratory Tests, SINTEF NBL as, Trondheim, Norway, STF25 A92026, 1993. 12. Chamberlain, G. A., Persaud, M. A., Wighus, R., and Drangsholt, G. Blast and Fire Engineering for Topside Structures. Test Programme F3, Confined Jet and Pool Fires, Final Report. SINTEF Report NBL A08102, 2008. 13. Chamberlain, G. A. An Experimental Study of Large-Scale Compartment Fires, Process Safety and Environmental Protection Tranactions Institute Chemical. Engineering, 72(4): 211-219, 1994. 14. Wighus, R., Lønvik, L. E., and Drangsholt, G. Fire on the Sea Surface: Thermal Load from Oil Spill Fires on the Sea Surface, SINTEF NBL as, Trondheim, Norway, SINTEF Report NBL A07127, 2007. 15. International Organization for Standardization. ISO 834-1:1999-Fire-Resistance Tests-Elements of Building Construction-Part 1: General Requirements, 1999. 16. Anon. Summary of the Large Scale Fire Tests in the Runehamar Tunnel in Norway, Conducted in Association with the UPTUN Research Program. TNO, Netherlands, 2003. 17. Standard Norge. EN 1363-2:1999, Fire Resistance Tests-Part 2: Alternative and Additional Procedures, Standard Norge, 1999.
305