7.3. Fires. acetone acetylene benzene

Transcription

1 7.3 Fires Introduction Fires (or, more generally, combustion processes) can be defined as an energy release following a chemical reaction between a fuel and a oxidant, the most common of which is the oxygen in the air. Fuels can be gaseous, liquid or solid, but flames always involve chemical phenomena between gaseous species. In order for a fire to develop fuel, oxidant (usually oxygen from the air), and energy (providing the ignition source) must be simultaneously present. Conversely, to prevent a fire from developing, the presence of one of these elements, either the fuel, the oxidant, or the ignition source, must be eliminated. To extinguish a fire it is necessary to remove either the fuel (for instance by interrupting the leak of a combustible liquid), the oxidant (for instance, by covering a pool of flammable liquid with foam), or the energy (for instance, by cooling a burning solid material with water). In order for a gaseous fuel/air mixture (or more generally a gaseous fuel/oxidant mixture) to sustain a combustion process, the fuel concentration must lie in a well-defined range, called the flammability range. The lower value of this range is called the Lower Flammability Limit (LFL), while the upper limit is called the Upper Flammability Limit (UFL). In other words, a gaseous fuel/oxidant mixture can ignite only when the gas concentration lies between the LFL and UFL values for the given fuel and oxidant and at given pressure and temperature values. s a rule of thumb, hydrocarbon/air Table 1. Flammability data for some fuels in air at ambient conditions. C st is the stoichiometric concentration Species LFL (% vol) LFL/C st UFL (% vol) UFL/C st MIE (mj) IT ( C) acetone acetylene benzene butane hexane ethane ethylene hydrogen methane pentane propane propylene VOLUME V / INSTRUMENTS 449

2 COMBUSTION ND DETONTION mixtures at ambient temperature and pressure show LFL values equal to about one half the stoichiometric concentration, while UFL is equal to about twice this concentration value. Flammability data for some species are summarized in Table 1. To ignite a gaseous fuel/air mixture the release of a very small amount of energy in a small volume is usually required, such as that provided by a spark. This energy allows the combustion reactions to begin and propagate to the bulk of gas. The minimum value of energy able to ignite a given fuel/air mixture is called MIE (Minimum Ignition Energy). By increasing the spark energy, the flammability range increases until it reaches an asymptotic value for high-spark energy values, as shown in Fig. 1. The LFL and UFL values are consequently measured using high values of the spark energy and they represent the asymptotic values of the flammability range shown in Fig. 1. Values of MIE for some species are summarized in Table 1. Flammability limits vary with temperature, as shown for the sake of example in Fig.. The flammability range on the temperature vs. concentration plane is bounded, apart from the two flammability limits, by the vapour saturation curve on one side (since at lower temperature values vapour condenses) and by the autoignition curve on the other side. The presence of a spark is not the only way to ignite a flammable mixture. When the mixture temperature exceeds a characteristic value (which is different for each fuel/air mixture), combustion reactions propagate throughout the gas mixture. This temperature limit is usually referred to as IT (uto Ignition Temperature). Some IT values are summarized in Table 1 and, for some linear paraffins, in Fig. 3. We can see that there is a discontinuity between n-butane and n-pentane; beyond which IT values remain almost unchanged. IT values strongly depend on the characteristics of the hot surface in contact with the flammable mixture, due to possible catalytic effects that are able to substantially modify the IT value. Reference values usually refer to laboratory flammability limits vapour concentration in air apparatus with clean surfaces; the presence of particular species or impurities on the surface (such as rust) can easily reduce the IT values by C. Pressure influences LFL and UFL values, but becomes significant only for pressure values very different from atmospheric pressure. The effect of significant pressure changes depends on the flammable mixture considered. For instance, when dealing with methane/air mixtures reducing the pressure below the atmospheric value increases the LFL and reduces the UFL until the two limits coincide and the mixture is no longer flammable. n increase of the pressure value with respect to the atmospheric value slightly reduces the LFL value and increases significantly the UFL value, thus enlarging significantly the flammability range. There are various methods for predicting fuel mixture flammability limits. The simplest is given by the Le Châtelier rule, stating that fuel mixture flammability limits are the result of a weighted average on the fuel mixture composition of the pure-species flammability limits: [1] LFL satured vapour-air mixtures mix = UFL = n mix n c c i LFL flammable mixtures LFL UFL FP T u IT temperature Fig.. Flammability range as a function of temperature. i= 1 i i= 1 i UFL autoignition i spark energy (mj) ignition limits IT ( C) MIE methane concentration in air (% volume) Fig. 1. Ignition limits for methane/air mixtures at ambient conditions as a function of spark energy (Drysdale, 000) number of carbon atoms Fig. 3. IT trend for paraffinic linear hydrocarbons. 450 ENCYCLOPEDI OF HYDROCRBONS

3 FIRES where c i represents the concentration (%vol) of species i-th in the mixture of fuels. The rule proposed by Le Châtelier is based on experimental evidences related to the lower flammability limit of light hydrocarbon mixtures. s a consequence, the main differences between the estimated values and the experimental measurements refer to the UFL values of mixtures of fuels with large differences in chemical behaviour as well as to situations where inert species (that is, species not participating in the combustion process) are present. different method, that tries from the start to consider the effect of the presence of inert species as well as the effect of the pressure and temperature changes on the mixture flammability limits, is the adiabatic flame temperature method. In this method, in order for a mixture to be able to sustain flame propagation, the minimum temperature reached by the burnt gases is about 1,500-1,600 K. conservative approach for the calculation of the burnt gas temperature is to assume that combustion reactions reach the equilibrium state without exchanging heat with the surroundings, reaching a temperature value called adiabatic flame temperature. For a given composition of the flammable mixture the value of the adiabatic flame temperature can be easily computed through thermodynamic information (that is, without involving the rate of the involved reactions). If the value of the adiabatic flame temperature is larger than 1,500-1,600 K the mixture is considered flammable. This approach provides reasonable estimations of LFL values, while its predictions are less reliable for UFL values. dding an inert species (i.e. a species that cannot burn, such as nitrogen, water, CO ) modifies the flammability characteristics of a fuel. Usually the presence of inert species considerably lowers the UFL without noticeably modifying the LFL. Increasing the amount of an inert species progressively narrows the flammability range until the two limits coincide and the mixture is no longer flammable. This behaviour defines a flammability region on diagrams indicating the relative amounts of fuel vs. inert species: mixtures inside this region are flammable; those outside are not. Different inert species modify the mixture s flammability characteristics in different ways (basically due to the different values of specific heat, leading to a different amount of heat being removed from the flame and consequently to different values of the adiabatic flame temperature) as shown in Fig. 4. It is clear that different inert species have different effects on the reduction of the flammability range. In the example shown in the figure, carbon dioxide has a more significant effect than nitrogen. Liquid or solid fuels emit vapour when heated (through liquid evaporation or solid pyrolysis, i.e. heavy molecules breaking down to produce volatile low molecular weight species) that can form a flammable mixture with air, in other words a mixture with composition between LFL and UFL values for the given species. This mixture can thus catch fire and fuel combustion can start. Heat build up is a necessary condition for the propagation of a fire. For gas combustion, the heat propane (% volume) flammable mixtures % air 100 (% propane % inert) Cst CO added inert gas (% volume) Fig. 4. Influence of inert gas addition on the flammability limits of propane/air mixture. C st is the stoichiometric composition line (Cardillo, 1998). released raises the temperature of the reactants until it reaches auto ignition temperature, thus sustaining combustion reactions. Instead, for liquid or solid fuels there is also a feedback process from the flame to the fuel that, through heating the fuel, leads to evaporation (when liquid) or pyrolysis (when solid) and therefore to an emission of flammable gases which sustain the flame itself. It is interesting to observe that since solid fuel pyrolysis usually requires more energy than liquid fuel evaporation, the temperature of solid fuels involved in a fire tends to be higher than the temperature of liquid fuels. While liquid fuel temperature is close to its normal boiling point, the surface temperature of solid fuels is typically about 400 C. solid or liquid fuel can emit flammable vapour through a combination of several different processes of melting, evaporation and decomposition. liquid fuel heated by a flame can either vaporize without modifying its composition (as light hydrocarbons do) or can partially decompose while evaporating (as heavier hydrocarbons do, for example). In both cases, the vapour coming from compounds with low molecular weight feed the flame. solid, instead, can sublimate or decompose directly generating flammable vapour, or it can melt (and possibly decompose at the same time) giving rise to a liquid that then can produce vapour following the mechanism described previously. For instance, light paraffinic hydrocarbons and waxes melt and then vaporize, while heavier waxes and thermoplastic polymers melt before decomposing and giving off vapour. Polyurethanes typically decompose producing both gaseous and liquid species that further decompose. Cellulose and many other thermosetting polymers decompose directly giving off gaseous species. In the case of solid or liquid fuels, it is not only necessary that vapour forms in order for combustion to take place, but also that the mixture of the vapour with the air has a concentration within the flammability N 50 VOLUME V / INSTRUMENTS 451

4 COMBUSTION ND DETONTION range. Very volatile liquids even at ambient temperature form enough vapour to overcome the LFL value close to the surface of the liquid. Less volatile liquids must be heated before they can release enough vapour to overcome the LFL threshold close to the surface and then, once ignition is present, catch fire. The minimum temperature value where a flammable vapour/air mixture can exist on the surface of a liquid (in other words, the minimum temperature value of the liquid fuel required in order for it to catch fire when ignited) is called the Flash Point (FP). Liquids at lower temperatures than their FP cannot catch fire when ignited for a short time, such us through a spark or a flame maintained for a short time. Obviously they can catch fire locally when a pilot flame is sustained long enough to heat the fuel, causing it to release enough vapour to overcome the LFL boundary locally. The flame generated in this way, thanks to the aforementioned feedback process, can then heat the bulk of the fuel, causing a large fire. The FP value is also shown in Fig., from which it can be seen that it corresponds to the minimum temperature value where vapour in the presence of liquid (on the saturation curve) reaches the LFL value. The same figure also shows the minimum temperature value, T u, beyond which the vapour concentration on the surface of the liquid is larger than UFL. t temperatures higher than T u the air/vapour mixture close to the surface of the liquid is no longer flammable and an ignition point close to the surface of the liquid cannot start a fire. It is important to remember that vapour concentration decreases when moving away from the surface of the liquid and consequently at a given distance from the surface of the liquid the vapour concentration will lie inside the flammability range. t this point, ignition is able to start a fire. These concepts are summarized in Fig. 5. Figure 5 shows a situation where the temperature of the liquid is lower than FP; in this case the vapour concentration close to the surface of the liquid is always lower than LFL and the liquid cannot catch fire following a short ignition at any distance from the surface. Figure 5 B represents a situation where the temperature of the liquid is higher than FP, but lower than T u ; in this case the vapour concentration near the surface of the liquid lies inside the flammability range and the liquid can catch fire when ignited close to its surface. Figure 5 C involves liquid temperature higher than T u ; in this case vapour concentration on the surface of the liquid is higher than UFL and the liquid cannot catch fire when ignited close to its surface, while it can when ignited far enough from the surface. It is clear that while FP values have many important practical implications for fire prevention, T u value has no practical relevance. few values of FP are summarized in Table. It is important to stress that all the aforementioned parameters depend substantially on the apparatus and the experimental procedures used for measuring them. It follows that when these values are to be used for safety purposes, suitable safety factors must be taken into consideration Flame structure: overview The various flame typologies can be classified in two large categories: premixed flames and diffusion flames. While in premixed flames fuel and oxidant are completely mixed and the flame propagates inside the mixture, in diffusion flames fuel and oxidant are separated and the flame can form only at the interface between the two species, where both fuel and oxidant are present and toward which both species diffuse. In other words, on one side of a premixed flame there are combustion products and on the other side the fuel/oxidant mixture; instead on one side of a diffusion flame there is fuel and on the other side oxidant. Bunsen burner is a typical example of a premixed flame, while a candle is a typical example of a diffusion flame. In a Bunsen burner air and fuel mix along the duct leading to the stabilized flame at the burner exit, while paraffinic vapours generated by a candle through the wick must diffuse towards air to create a flammable mixture and consequently a flame. In premixed flames the characteristic time of the combustion phenomenon is related to the transport rate of the heat and radical species (in particular, hydrogen atoms) from the flame towards the unburned gases, while in diffusion flames the characteristic time of the combustion phenomenon is related to the transport rate of LFL LFL UFL LFL UFL B C Fig. 5. Schematic representation of the effect of liquid temperature on flammability. 45 ENCYCLOPEDI OF HYDROCRBONS

5 FIRES Table. Flash Point, FP, in closed cup for some species with air at ambient pressure reactant concentration preheating region flame region T b acetone benzene butane hexane ethane ethanol Species FP ( C) concentration, temperature unburned gases (fuel oxidant) temperature T o T i burned gases concentration of intermediate products ethylene 11 methanol 10 pentane propane propylene toluene 4 the molecular species towards the fuel/oxidant interface where the flame is located. In both cases the transport rate depends on the fluid dynamic conditions and increases strongly switching from laminar to turbulent regime. Fires arising from industrial accidents almost always involve diffusion flames. The only relevant exception is the atmospheric dispersion of a cloud of flammable vapours mixing with air before finding an ignition source. In this case flame front propagation proceeds in the fuel/air mixture and can generate, apart from a fire, an explosion called UVCE (Unconfined Vapour Cloud Explosion). Premixed flames The typical structure of a premixed flame is shown in Fig. 6, where the trends of temperature (from the unburned gas value, T o, to the burned gas value, T b ), the concentration of reactant (fuel and oxidant, from their stoichiometric mixture values in the unburned mixture to zero in the burned gas region) and the concentration of intermediate products (whose concentration reaches a maximum in the flame). Three regions can be identified in the flame structure shown in Fig. 6: a first preheating region of the unburned gases where the temperature, thanks to heat transport from the high temperature burned gases, increases from the value T o of the fuel/oxidant mixture to a value T i, arbitrarily assumed to be the ignition point of the considered mixture and beyond which combustion reactions proceed at a high rate; a second reaction region where most of the combustion reactions occur and which represents the visible flame region (for hydrocarbon/air mixtures at ambient pressure the depth of this region is about one mm); finally, a high Fig. 6. Temperature and concentration profiles in a premixed flat flame. temperature T b region of the burned gases where radical recombination reactions occur and equilibrium conditions are attained. From this region, hot burned gases exchange heat with the environment and cool down. This diagram highlights the existence of a preheating region and the importance of the heat transport processes for the propagation of a premixed flame. ctually, the transport of radical species (particularly hydrogen radical transport due to its high diffusivity) from flame to unburned gases also helps set off the radical reactions of combustion. When the flammable mixture is at rest the flame front velocity with respect to unburned gases depends only on the physico-chemical properties of the flammable mixture, apart from its composition, temperature and pressure. If the flame propagates in a quiescent mixture of unburned gases the heat and mass transport phenomena are determined by the overall transport coefficients (mass and heat diffusivity) as predicted by the Fick and Fourier laws where the flame propagates at a given velocity that, for hydrocarbon/air mixtures slightly richer than the stoichiometric value, is usually equal to about 0.5 m/s. few values of this burning velocity are summarized in Table 3, from which it can be seen that the more relevant exceptions to the aforementioned value refer to very reactive fuels, such as hydrogen or acetylene. n order-of-magnitude estimation of the burning velocity can be done taking into account only the effect of the preheating of unburned gas by an infinite flat flame, for which the energy balance can be recast as: [] k dt ruc dt P Q = 0 dx dx where k represents the thermal conductivity, T the temperature, x the coordinate normal to the flame plane, r the density, u the gas velocity, C P the specific heat and Q the heat release rate per unit volume due to combustion reactions. If the overall reaction rate strongly depends on temperature (in other words, if Q exp( E RT) and the activation energy, E, is very high) it is possible to ignore the combustion reactions in the preheating region (note VOLUME V / INSTRUMENTS 453

6 COMBUSTION ND DETONTION Table 3. Maximum values of the laminar burning velocity, S o, of some gaseous fuels Fuel S o (m s 1 ) acetylene 1.58 benzene 0.6 butane 0.50 butene 0.57 cyclohexane 0.5 heptane 0.5 hexane 0.5 ethane 0.53 ethylene 0.83 hydrogen 3.50 methane 0.45 pentane 0.5 propane 0.5 propylene 0.66 that ignition temperature, T i, is defined as shown in Fig. 6) and to analyse the two regions (i.e. the preheating and reaction regions) separately to obtain two relations for the temperature gradient that must be equal where T T i. Following this approach, due to Zeldovich, Frank-Kamenetski and Semenov, the following expression for the laminar burning velocity, S o, can be deduced: kq [3] So = ( ocp ) r ( Tb To ) where Q is the average rate of heat release (over temperature). This relation, even approximated, gives the correct dependence, confirmed by experimental results, of the laminar burning velocity on the main physico-chemical parameters (k, r, C P, T b and T 0 ). Moreover, it predicts a dependence on the overall reaction rate, involved in the Q term, as S o 1 exp( 1 1 E RT ) exp( E RT). If the fluid dynamic conditions of the unburned gases are turbulent, the flame front speed no longer depends solely on the physico-chemical characteristics of the flammable mixture, but also on the fluid dynamics that determines the turbulence level of the flammable mixture in which the flame propagates. The turbulent flow of the unburned gases leads to an increase of the heat and mass transport rates from the flame to the unburned gases since the efficient mass and heat transport coefficients (turbulent mass and heat diffusivity) are much larger than the analogous molecular coefficients. To predict the possibility of a flame extinction it is necessary to take into account heat loss from the flame. This heat loss leads to a reduction of the flame temperature from the ideal adiabatic value, T b, to a lower value, T F. This induces a strong reduction in the heat release rate due to the exponential relation between temperature and reaction rate, and hence between temperature and heat release rate. flame goes out when the heat loss rate due to heat exchange with the environment is larger than the heat release rate due to combustion reactions. For adiabatic flames, through an approximate energy balance (assuming constant physico-chemical parameters) across the flame, the burned gas temperature can be related to the combustion heat, Dh comb, as: [4] S rc T T S r h ( ) = o o P b o o o comb When heat can be exchanged with the environment the heat loss term, U(k, T F ), must be included in the energy balance; this term depends on thermal conductivity as well as on flame temperature and leads to the relation: [5] S rc T T S h U k T o o P( F o) = r (, ) o o comb F Keeping in mind that, for laminar flow conditions, S o exp( E RT) or, in other terms, S o exp( E RT), and that C P (T b T o ) Dh comb, the following relation can be deduced: [6] UkT (, F) = roct P ( b TF) exp( E RT) whose trend is shown in Fig. 7 ( is a proportionality constant). The value of the adiabatic flame temperature, T b, obviously corresponds to a nil value of the heat loss term. The heat which through the convection mechanism is exchanged (per unit of flame surface) with the duct walls (assumed to be of diameter D and full of unburned gases) where a flame of depth d propagates is given by: ( ) π [7] q = h T T U(k,T F ), q conv conv F o q conv πd b 1 Dδ 4 b U(k,T F ) flame temperature Fig. 7. Relation between flame temperature, heat loss by the flame itself and heat exchanged with the walls of a duct of various diameters (D 1 D D Q ). a a 1 T b D Q D D ENCYCLOPEDI OF HYDROCRBONS

7 FIRES where h is the heat transfer coefficient, related to the Nusselt number, Nu, by the relation h Nu k D 3.65 k D. Moreover, the flame front depth can be approximated by the relation d k r o S o C P to give: [8] q conv ( ) k T T F DS o showing that the convection heat rate to the duct walls per unit of flame surface increases when the duct diameter decreases, but decreases when flame temperature increases due to the exponential dependence of the burning velocity on temperature. The trend of convection heat, for various duct diameters, is also shown in Fig. 7; the intersections between the curves representing heat release rate and heat removal rate represent the flame temperature values for which the flame can propagate into the duct. It can be easily seen that only the intersections labelled a represent stable propagation points since, following a flame temperature increase, the heat removed by conduction is greater than that released by the flame, leading to the cooling of the flame and a return to the initial conditions; the same is true for a flame temperature decrease, while the opposite is true for the points labelled b. More interesting is how these intersections move when the duct diameter changes. s can be seen from Fig. 7, there is a critical value of the duct diameter, D Q, beyond which the two curves can no longer intersect and consequently flame propagation is not possible. lthough this analysis has been greatly simplified (in particular, the radical quenching on the walls has not been considered) the main trends have been confirmed by experimental results, with particular reference to the existence of a minimum diameter for flame propagation, which is characteristic of any fuel/oxidant mixture, beyond which a premixed flame cannot propagate into a duct. The maximum value of the duct diameter allowing for flame suppression is called the quenching diameter and plays an important role in the design of flame arresters which protect industrial units from flames propagating into ducts. nalogous to the quenching diameter, there is also a critical width of a slot able to suppress flames. The quenching diameter, D Q, can be related to the quenching width of a slot, d Q, by the empirical relation d Q 0.65 D Q. Values (in mm) of the quenching width of a slot at ambient temperature and pressure for stoichiometric fuel/air mixtures are: 0.50 for hydrogen to 1.87 for benzene, 1.5 for ethylene,.16 for methane,.07 for pentane and 1.75 for propane. o Diffusion flames The main characteristic of diffusion flames, as previously mentioned, is that fuel and oxidant are initially separated and combustion proceeds where the gases mix. The simplest accidental situation of industrial relevance involving a diffusion flame is a et of flammable gas. When a gaseous et enters an atmosphere at rest it entrains some air that, mixing with the fuel, allows a flame to form whose front is localized where fuel and oxidant meet each other. t low et velocity, the flame is laminar and its length is roughly proportional to the square root of the volumetric fuel flow rate, but when the et velocity increases the flame becomes turbulent and its length no longer depends on the fuel flow rate. Further increasing the et velocity, the amount of entrained air close to the flame tip becomes too large to allow flame attachment close to the et exit and flame blows off. Depending on the fuel flow velocity, the flame that is formed can be dominated by buoyancy (for low velocity) or by momentum (for high velocity). The dimensionless number representing the ratio of the momentum to buoyancy is the Froude number, usually given by Fr u gd, where u represents the fuel velocity, g the gravitational acceleration and D a characteristic dimension of the flame, for instance the diameter of the fuel et opening or the diameter of the pool of liquid where fuel vapour is produced. Turbulent flames generated by ets of gaseous fuel are characterized by high Fr values. Flames originating from the combustion of condensed fuels (liquids or solids) are usually characterized by buoyancy since the fuel vapour velocity is low. This velocity can be estimated from the combustion heat release rate, Q, for a pool fire of diameter D as Q Dh comb r u pd 4 it follows that Fr Q D 5. When the characteristic dimension of the fuel vapour is lower than about 0.05 m the flame is laminar, while for dimensions larger than about 0.3 m the flame is turbulent. Through the Froude number several other parameters can be defined which are used to relate some characteristic diffusion fire parameters, such as flame length. One of these parameters is a dimensionless thermal power which is proportional to the square root of the Froude number: 131 Q D 5 Q* rc P T g and correlates flame length (dimensionless through its characteristic dimension) as shown for the sake of example in Fig. 8. The regions on the left-hand side of this figure, labelled I and II, are characterized by low Q* values and correspond to buoyancy flames, such as those arising from the combustion of solid or liquid fuel. The region on the right-hand side of the same figure, labelled V, is characterized by high Q* values and corresponds to et flames, such as those originated by fully turbulent ets of gaseous fuels. Indoor fires (such as fires inside buildings or industrial warehouses) are usually characterized by Q* 5, while industrial accidents can involve both pool fires of relevant dimensions characterized by Q* 1, as well as fully turbulent et fires characterized by Q* In region V of Fig. 8 flame length is roughly constant. In regions I and II of the same figure flame length varies quickly with Q*. The shape of these last flames is much less stable than that of momentum-dominated flames and is more sensitive to environmental parameters, such as wind. The structure of these flames usually involves three different regions: the first (persistent flame) close to the VOLUME V / INSTRUMENTS 455

8 COMBUSTION ND DETONTION H/D I II III Q* Fig. 8. Diffusive flame length, H, made dimensionless with respect to a characteristic dimension, D, as a function of the dimensionless thermal power, Q* (Drysdale, 000). IV V regions, is very important for evaluating the possibility that neighbouring plant units might be involved in the fire, either by direct flame impingement or by heat radiation from the flame. Dimensionless heat power Q* represents a useful parameter for predicting flame height, as shown in Fig. 9 where various experimental data are given as a function of this parameter. For high values of the ratio of height to a characteristic dimension of the flame, H D 6, it has been found that H D (Q D 5 ) 0.4. For lower values of the ratio of height to a characteristic dimension of the flame, 10 H D 3, the well-known Thomas relation gives H D (Q D 5 ) For even lower values, H D, the relation between H D and Q/D 5/ becomes almost linear, while for H D 0. it becomes proportional to the square of the dimensionless heat power. These different correlations between flame height and dimensionless heat power agree with the qualitative trends summarized in Fig. 8 and have already been discussed. fuel where the flow rate of the burned gases increases with height; the second (intermittent flame) where the flow rate of the burned gases is roughly constant; the third (buoyant plume) is characterized by a reduction of both velocity and temperature of the burned gases as the height increases. From a qualitative point of view, the hot buoyant plume rises due to buoyancy forces caused by the density difference between environmental air and the hot burned gases. s the plume rises higher, its movement is slowed down by cold air entrainment which reduces the temperature of the plume and consequently its buoyancy. t a given distance from the flame, buoyancy vanishes and the plume can rise no higher. The classification of the movement of a plume is extremely important, for instance for positioning indoor fire detectors, since the hot plume must reach the detectors in order for the fire to be detected. nalogously, the classification of the geometry of the flame zone, both in the persistent and intermittent Industrial fires The consequences of industrial fires are basically due to the thermal effects of the flames, as well as to indirect effects due to the atmospheric dispersion of toxic by-products caused by incomplete fuel combustion. These by-products can be hazardous both for people and the environment, and they often represent the maor source of risk from a fire. In particular, for indoor fires (that is, fires developing inside buildings or warehouses) smoke and carbon monoxide (inevitable when the fire involves organic material, is indoors, and where there is a limited air supply) often represent the main source, either directly or indirectly, of death for the people inside the structure. It follows that an analysis of the effects of fire must also take into account the dispersion of toxic combustion by-products, both inside the structure where the fire occurs and in the surrounding environment. This means that for buildings and Fig. 9. Diffusive flame length, H, made dimensionless with respect to a characteristic dimension, D, as a function of the dimensionless thermal power, Q* or, analogously, of the group Q/D 5/ (Drysdale, 000). H/D Q/D 5/ (kwm 5/ ) Q* 456 ENCYCLOPEDI OF HYDROCRBONS

9 FIRES warehouses emergency exits must be provided that permit people to safely leave the building. complete discussion of these topics is outside the scope of this paper and consequently only thermal effects will be discussed here. Thermal effects can be either direct, when the flame impinges the target, or indirect, when the flame radiates energy to a target outside the flame itself. Direct effects are usually lethal for people and cause serious damage to structures and plant units, especially when the flame impingement is prolonged. In this case, the fire can trigger other accidents due to the collapse of structures or plant units, creating a domino effect. The effect of heat radiation on people depends on the specific power reaching the target as well as on the time exposure. Significant damage to people is expected only for high radiation power, since lower radiation values are dangerous only for long exposure times which allow people to escape. From this point of view, the most dangerous fires for people are those with very high heat radiation and of very short duration, such as fireballs which will be discussed below. On the contrary, assuming unlimited exposure it is possible to define threshold values for damage to plant units, as summarized in Table 4. In this table, for the sake of example, the threshold exposure limits for people related to various exposure times are also given. To quantify thermal effects of an industrial fire it is, therefore, necessary to define the flame geometry and heat radiation on the surroundings. few methodologies to estimate such effects will be discussed after a general introduction of the various typologies and sources of industrial fires. Typologies and sources of industrial fires The source of a maor industrial fire is usually the discharge of a flammable fluid from a plant unit (tank, drum, reactor, apparatus, pipe, etc.). The only relevant exception is the combustion of a solid material stored in the plant. Table 4. Threshold values of specific radiation power for various types of damage Threshold radiation power (kw/m ) Damage 37.5 damage to process equipment 5.0 non-piloted ignition of wood 1.5 damage to or melting of plastic pain tolerated by people for only a few seconds pain tolerated by people for up to 1 minute with protective clothing 4.5 causes pain in 0 seconds 1.6 no effects even for long exposures The discharge of a flammable fluid can originate either from the catastrophic collapse of an apparatus or from a leak, as summarized in the flow diagram in Fig. 10. Events following the catastrophic collapse of a plant unit depend on the phase of the fluid released into the environment. When a flammable gas is released, a cloud forms which disperses in the atmosphere and is progressively diluted with air until a significant amount of fuel/air mixture is created whose concentration is inside the flammability range. In this case, after ignition, there is a first phase characterized by premixed flame propagation, followed by a second phase involving gas concentrations larger than the UFL in a diffusion flame. When the characteristic times for combustion energy release are not fast enough to generate an explosion, a flash fire arises. The main consequences of this phenomenon are the direct impingement of people and structures in the region where the flame travels, as well as heat radiation to the surroundings. If the collapsing unit also contains a liquid phase at ambient temperature two situations can arise depending on the value of the normal boiling point of the substance involved with respect to the ambient temperature. When the temperature of the liquid (which is assumed to be equal to the ambient temperature) is lower than its normal boiling point, the liquid is subcooled and the characteristic times related to liquid evaporation are usually quite high. The vapour formed from the slow evaporation of the liquid, when no immediate ignition is provided, disperses in the atmosphere and can create the aforementioned phenomena, while the liquid phase can give rise to a pool fire. The main consequence of this phenomenon is the thermal radiation of the surroundings. When, instead, the temperature of the liquid is higher than its normal boiling point, the liquid is overheated (and therefore in non-equilibrium conditions) and consequently the characteristic times related to liquid evaporation (either complete or partial) are much shorter and a flash can occur. The vapour and aerosol formed from the fast evaporation of the liquid, when no ignition is present, disperse into the atmosphere and can lead to the aforementioned phenomena. nalogously, the rained-out liquid, even if usually a small quantity since most of the nonevaporated liquid remains trapped in the cloud as an aerosol, can give rise to a pool fire. In the presence of an immediate ignition a fireball can arise. When a large mass of flammable liquid evaporates very quickly a vapour cloud forms, only a small part of which, around the outside of the cloud, initially has a concentration within the flammability range. The presence of an immediate ignition leads to the formation of a flame on the cloud surface. Since the fuel cloud is inside the flame, a diffusion flame locates on the surface of the cloud itself. s combustion proceeds, the fuel cloud on the one hand burns away and on the other hand heats up; its density is consequently reduced and so the burning fuel cloud moves upward due to buoyancy. The main consequence of this phenomenon is the very high heat radiation to the surroundings. The discharge of process fluid can be also due to a non-catastrophic failure, such as a leak from a flange or a VOLUME V / INSTRUMENTS 457

10 COMBUSTION ND DETONTION failure of an apparatus loss from an apparatus flammable solid containing flammable liquid containing flammable gas flammable gas flammable liquid ignition 1 delayed ignition ignition T eb T L T eb T L T eb T L T eb T L fire ignition delayed ignition ignition atmospheric dispersion et-fire 1 flash pool-fire evaporation ( flash) 3 fireball ignition gas aerosol liquid 3 liquid gas aerosol flash-fire Fig. 10. Flow diagram of fire generation in an industrial plant or warehouse (T L is the temperature of the liquid and T eb is its boiling temperature). pipe crack. Depending on the phase (gas, liquid or twophase) of the released fluid the aforementioned dispersion phenomena occur. The only difference is the immediate ignition of a gaseous (or two-phase) et resulting in a et fire whose main consequences are, as always, thermal effects (heating of the walls of a plant unit impinged by the et or radiation on the surroundings). Pool fires pool fire can be defined as a turbulent diffusion flame over the horizontal surface of a liquid fuel that evaporates so slowly that its initial momentum is almost nil. fundamental characteristic of this kind of fire is the feedback from flame to fuel since the heat transfer from the flame to the liquid pool influences (or even controls) the fuel evaporation rate which in turn feeds the flame. The main variables influencing a pool fire dimensions are the pool diameter, the temperature of the released fluid and the wind velocity. The characteristics of a fire for pool diameters lower than one metre are quite different from larger pool diameters, which are the norm for industrial fires. The temperature of the released fluid determines whether part of the liquid gives rise to a flash fire and the influence, with respect to flame radiation, of heat transfer from the soil. This last phenomenon can be relevant for cryogenic fluids where the soil characteristics can also influence the evaporation rate of the liquid. Finally, wind velocity plays a relevant role on the flame height and tilt, as well as on the base flame drag in the direction of the wind. The evaluation of pool fire consequences requires, as previously mentioned, a flame model in terms of geometry and heat power radiation. model of this type usually makes use of several submodels, discussed below, for each of the considered phenomena: burning rate of the liquid fuel (equal to the fuel evaporation rate feeding the flame), pool diameter, flame geometry, view factor (which represents the amount of energy, emitted from the flame that hits the surface of a given target, to suface unit), atmospheric transmissivity (that is, the fraction of the energy emitted from the flame that is not absorbed by the atmosphere) and thermal power emitted by the flame. Pool dimension (D) can be fixed by a confinement, such as a dike around a reservoir. Even when there is a confinement, the maximum pool dimension can be lower than that of the dike when the fuel loss is small. The fuel discharge can be considered instantaneous or continuous depending on the time of the release, t R, with respect to the characteristic combustion time, t C. The latter can be estimated as the ratio of a vertical characteristic dimension of the pool to the linear burning rate, y in m/s. The burning rate can be expressed as a specific massive value with respect to the pool surface area kg/(m s) or as a linear value (m/s), seeing that the two values are related by the liquid fuel density (kg/m 3 ). When there is an instantaneous release, the unconfined pool dimensions depend on the depth limit of the liquid on a given soil type, while for a continuous release the pool expands until the fuel burning rate equals the fuel release rate, m L : [9] m = y πd L 4 ctually, the maximum value of the pool diameter is larger than the value calculated using the previous relation in that during the period of time from the beginning of the discharge to the moment when the system reaches equilibrium the fuel release rate is lower than its maximum value. Therefore, some released liquid accumulates leading to a small increase of the maximum 458 ENCYCLOPEDI OF HYDROCRBONS

11 FIRES Table 5. Considering that these experimental measurements are highly uncertain, these values can be represented by the following relation: radiation D important for cryogenic fluids soil conduction D flame conduction D Fig. 11. Thermal exchanges between liquid fuel pool and environment determining burning rate. exchange with the environment D important for small diameters or large DT pool diameter (equal to about 4 ), which then reduces and reaches the equilibrium value. The burning rate is the result of a balance of the energy exchanged between the liquid pool and the environment and the energy required to vaporize the liquid, as shown in Fig. 11. We can see that all the contributions depend on the square of the pool dimension except for the environmental heat transport rate, which is consequently relevant only for small pool diameters (less than one metre) or for high temperature differences between the fuel and the environment (impossible for liquid fuels with low boiling points, but quite common for solid fuels that, as mentioned previously, must reach elevated temperatures in order to release combustible gases). This, together with the various fluid dynamic regimes of the flame which depend on the pool dimension, explains the burning velocity trend shown in Fig. 1 where it can be seen that, for pool dimensions larger than 1 m, the burning velocity is almost constant. This asymptotic value is summarized for a few fuels in hcomb [10] y = Tb h + C dt ev T P a where Dh comb and Dh ev are the combustion and evaporation heat, respectively. ssuming an average density for liquid fuels equal to about 787 kg/m 3 the previous relation provides an alternative equation for computing the specific mass burning rate: H comb [11] m = 10 3 b Tb H + C dt ev T P a Estimating the burning rate on water is more complex. For subcooled liquid fuels this rate is quite similar to the rate on soil, while for fuels where heat exchange with soil or water plays a relevant role in determining the evaporation rate (such as LNG or LPG), the burning rate on water can be much higher than on soil (for instance, by a factor of about for LPG or 3 for LNG), as shown in Table 5. The flame geometry of a pool fire is not well defined, also because it is characterized by a periodic behaviour with a return time of about 1 s, where large bright hot zones reach the external flame surface breaking the smoke shield. Consequently, the values estimated from the available relations are time-averaged values, as are the radiation values estimated using these flame geometries. There are numerous correlations for calculating the maximum height of the visible part of the flame, the bestknown being the Thomas relation: 061. H m b [1] = 4 D r gd Wind affects flame geometry in diverse ways: it causes a tilt of the axis of the cylinder on which the flame has been modelled; it modifies the flame height; and it tends to drag the base of the pool flame in the direction of the wind. The relations that represent the influence of 0 burning velocity (mm/min) laminar gasoline kerosene fuel oil diesel transition turbulent ,000 3,000 pool diameter (cm) Fig. 1. Trend of the burning rate of pure liquid fuels vs. pool dimension. VOLUME V / INSTRUMENTS 459

12 COMBUSTION ND DETONTION Table 5. Burning rate for some liquid fuels Species on soil Burning rate (m/s) 10 4 on water benzene butane butene cyclohexane cyclopentane heptane hexane ethane 1..7 ethylene hydrogen isobutane isohexane isopentane methane nonane octane pentane propane propylene toluene xylene dimensions can be related to the Froude number (based on wind velocity) as: D' [14] = Fr D D being the pool dimension in the wind direction. Even a slow wind speed can significantly increase dimension of the burning pool in the wind direction. This is more important for evaluating flame impingement of surrounding plant units rather than for estimating fire radiation on the environment. nalogously, the flame tilt ÿ with respect to the vertical direction due to the wind can also be estimated on the basis of the Froude and Reynolds number values as: tanθ.. [15]. cosθ = Fr Re where the Reynolds number is based on air properties, wind velocity at 10 m and pool diameter. Knowing the flame geometry is also essential for estimating the radiation intensity reaching a given target near the flame since the radiation from the flame to a target depends on the flame shape, as well as on the distance between the flame and the target and their relative orientation. Not all the energy radiated from a flame can be intercepted by the target; the fraction of energy radiated from the flame that is intercepted by the target (per unit of target surface) is called the view factor, F, and can be computed using the relation: cos icos [16] F = β β di πx 1 where b i and b are the angles between normal direction and the straight line from the infinitesimal flame radiant element (whose surface area is equal to d i ) and the target, while X is the distance between the infinitesimal flame radiant element and the target, as shown in Fig. 13. Integration over the whole flame surface can be carried out analytically for a few simple geometries. The relations that give the F value can be deduced for several simple geometries and are available both as equations and tables or diagrams. Some examples of maximum view wind on flame height are usually given in the same form as the Thomas relation: B H [13] D b C = u mgd (*) r where, however, a dimensionless wind velocity, u* u (gm b D r ) 1 3, is involved. Various values for the parameters of this equation have been proposed. Originally, Thomas proposed the values 55, B 0.67 and C 0.1. More recently, Moorhouse proposed the values 6., B 0.54 and C 0.044, which can be used when u* u 10 (gm b D r ) (when calculating u* the wind velocity at a height of 10 m must be used). Flame drag in the wind direction results in a larger pool length in the wind direction than in the direction perpendicular to the wind, leading to an elliptical flame base. The ratio between the two characteristic pool d n b X n b i i i d i Fig. 13. Schematic representation of the view factor. 460 ENCYCLOPEDI OF HYDROCRBONS