Twenty-first Symposium (International) on Combustion/The Combustion Institute, 1986/pp. 325-333
V O L A T I L I T Y M O D E L FOR COAL D U S T FLAME P R O P A G A T I O N A N D EXTINGUISHMENT MARTIN HERTZBERG. ISAAC A. ZLOCHOWER, AND KENNETH L. CASHDOLLAR
Pittsburgh Research Center, Bureau of Mines U.S. Department of the Interior Pittsburgh, Penmyh,ania 15236
A theoretical analysis is presented for the propagation and extinguishment of coal dust flames and of dust and gas flames containing inhibitor powders. The analysis is based on the established inechanislns for homogeneous flame propagation and the well known concept of a constant lilnit flame temperature for a given class of homogeneous fuels. The analysis is expanded to phase-heterogeneous systems such as coal dust by means of a volatility model. The analysis includes the singly heterogeneous system of a solid fuel dust in air; the singly heterogeneous solid inhibitor dust in a homogeneous fuel-air flame, and the doubly heterogeneous system consisting of a solid fuel and inhibitor dust mixture in air. The data for measured explosion pressures, flammability limits, and extinguishant requirements for heterogeneous systems are shown to be consistent with the established mechanisms and processes for homogeneous flame propagation provided that one adds an additional process: the heating and devolatilization of the solid fuel or inhibitor. The limitations on the rates of devolatilization of the solid particles become rate controlling at high burning velocities, at high dust loadings, and for large particle sizes. Devolatization rates are controlled bv the intrinsic devolatilization rate constant for the solid fuel or inhibitor and the effective heating flux of the approaching flame front. The effective yield of volatiles is a function of those factors, the decomposition chemistry, and the time available for devolatilization. The fraction of the total volatiles that can be generated in the time available is the [3-factor, and it determines the effective yield of fuel or inhibitor that participates in the flame propagation process. The data for explosion pressures, flammability limits. and extinguishant requirements are readily understood in terms of those [3-factors.
Introduction In several previous studies, data were rep o r t e d on the flammability limits and ignitabilitv b e h a v i o r o f coal dusts a n d their d e p e n d e n c e on coal rank. particle size. o x y g e n content, and pressure. '.... A d d m o n a l studies on the effectiveness o f inhibiting dusts in e x t i n g u i s h i n g both coal dust-air and m e t h a n e - a i r explosions have also been r e p o r t e d . 56 T h i s p a p e r presents a theoretical analysis o f those data. T h e analysis or "'model" is based on the established concepts and m e c h a n i s m s for h o m o g e n e o u s flame p r o p a g a t i o n , which are e x p a n d e d to h e t e r o g e neous svstems by the inclusion o f one additional process: the h e a t i n g and devolatilization o f the solid coal dust o r inhibitor particles. T h e h e t e r o g e n e o u s systems to be c o n s i d e r e d can be c a t e g o r i z e d as singly h e t e r o g e n e o u s or doubly h e t e r o g e n e o u s . A solid fuel dust such as coal d i s p e r s e d in air is singl.v h e t e r o g e n e o u s . Similarly, a solid inhibitor dust dispersed in an
otherwise h o m o g e n e o u s fuel-air m i x t u r e is also singly h e t e r o g e n e o u s ; but a coal dust and inhibitor dust m i x t u r e d i s p e r s e d in air is doubly h e t e r o g e n e o u s . All o f those systems are enc o u n t e r e d in c o n s i d e r i n g the p r o b l e m s o f prev e n t i o n and e x t i n g u i s h m e n t o f explosions in the fossil fuel m i n i n g i n d u s t r y 7 and in industries that m a n u f a c t u r e , transport, store, and use f l a m m a b l e substances, p u l v e r i z e d fuels, agricultural ~roducts, o r g a n i c powders, and metal dusts, s'" T h e analysis o r m o d e l to be p r e s e n t e d starts ~dth h o m o g e n e o u s flames a n d is based on the established concept o f constant limit flame t e m p e r a t u r e s . T h e limit flame t e m p e r a t u r e may be d e f i n e d as the calculated adiabatic e q u i l i b r i u m t e m p e r a t u r e for the m e a s u r e d limit composition, a n d it has b e e n shown that its value is relatively constant for a given class o f gaseous h y d r o c a r b o n fuels. 1~ Within a given h o m o l o g o u s series o f h y d r o c a r b o n - l i k e fuels (alkanes, alkenes, alkynes, amines, alco-
COAL COMBUSTION FUEL EQUIVALENCE RATIO
hols, aromatics, etc.), the lean limit compositions contain a relatively constant mixture heat of combustion, and that is reflected in a constant adiabatic flame temperature for the limit compositions. 13
Adiabatic Flame Temperature Calculations 1,600
The adiabatic flame temperature calculations were made with the NASA-Lewis computer code, CEC 80] 4 using the best available thermodynamic properties taken from the J A N A F tables. ]5 The calculations can be made u n d e r constant pressure constraints (isobaric, HP) or u n d e r constant volume constraints (isochoric, UV). Naturally, there are no kinetic limitations in such equilibrium calculations so an infinite time is assumed to be available for the chemical transformation; nor are there spatial gradients in any of the thermodynamic variables. Once the adiabatic flame temperature is calculated as a function of composition, the choice of a limit flame temperature is sufficient to determine the predicted limit composition. It is implicit that final equilibrium is attained by the reactants at the limit composition, or that the departure from equilibrium conditions is constant within a given class of fuels. Since most of the available data on flammability limit compositions are obtained at constant pressure] 6''' it is the constant pressure adiabatic flame temperature that is usually specified as the limit flame temperature.
9 8 s
5 4 3 2
ij Experiment !/----Spark . . . . . . 500-d igmfor ---- ~ I~O00-J ignito,
~"~ ~ ~
METHANE~ vol pcl
Fro. 1. --Calculated adiabatic flame temperatures for constant pressure combustion of methane in air (A) and calculated adiabatic explosion pressures for constant volume combustion compared with measurements (B).
A similar analysis for rich mixtures gives a value of 18 to 19 pct for the rich limit, T h e r m o d y n a m i c calculations for adiabatic , substantially higher than the previous literature flame temperatures of methane in air are value of 15 pct. ]&17 T h e i8 to 19 pct value shown in Fig. 1. T h e solid curve in Fig. 1A is the corresponds to a limit flame temperature of adiabatic flame temperature for constant pres1,400 to 1,500 K, a n d the fact that both limits sure (1 atm) combustion (HP calculation) as a now give nearly the same value supports the function of m e t h a n e concentration in air. I n reasonableness of the limit flame temperature Fig. 1B, the solid curve is a constant volume concept. calculation (UV), which is compared with the measured explosion pressure values. T h e comparison is made for different ignition sources: Polyethylene Dust one spark source a n d two pyrotechnic ignitors (500 and 1,000 J) whose characteristics are The extrapolation from methane to coal dust described elsewhere. 4 A careful analysis of the volatiles is too large a step to take in view of the methane data together with an analysis of the uncertainties associated with the composition criterion chosen for determining the limit and a m o u n t of the volatiles emitted from coal composition is given elsewhere. 4 T h e c u r r e n t during its pyrolysis prior to combustion. T h e best value for the lean limit is 4.9 pct, 4,}s shown intermediate case of polyethylene is simpler as the vertical arrow. It corresponds to a because the polymer normally volatilizes comconstant pressure adiabatic flame temperature pletely u p o n rapid pyrolysis. T h e calculated of 1,450 K, which is therefore the limit flame adiabatic (HP) flame temperatures for polytemperature for methane. ethylene in air are shown in Fig. 2 as a function
VOLATILITY MODEL FOR COAL DUST FLAMES I
2,400 [ - ~
': ii 1,200
POLYETHYLENE CONCENTRATION, g/m 3
FIG. 2. --Calculated adiabatic flame temperatures for constant pressure combustion of polyethylene at three initial pressures with arrows indicating the measured lean limits of flammability at those pressures. of polyethylene concentration for three initial mixture pressures. T h e m e a s u r e d lean limits 4'27 at those pressures are shown as vertical arrows. T h e average of the three limit flame temperatures is 1,380 K, lower than the value for methane. T h e solid polyethylene polymer, (C2H4)n, is synthesized from its m o n o m e r ethylene gas, C2H4. However, the pyrolysis and devolatilization of polyethylene generates very little monomer; it pyrolyses into much higher molecular weight fragments that contain fewer unsaturated bonds than the monomer, and, accordingly, those volatiles are intermediate between the fully saturated alkanes and the fully unsaturated alkene m o n o m e r . In Fig. 3, the calculated limit flame temperatures for 2,O00
both the alkenes and alkanes are shown. It is not surprising that the limit flame temperature for the polyethylene volatiles is 1,380 K, intermediate between the fully unsaturated ethylene value and the fully saturated alkane values.
Summary for Homogeneous Gases
T h e full range o f limit flame temperatures for homogeneous gas mixtures is shown in Fig. 3, where the values are plotted as a function of carbon n u m b e r for the homologous series of alkanes, alkenes (olefins), alkynes (acetylenes), benzene derivatives, amines, a n d halogenated hydrocarbons. T h e lean limit composition data were taken mainly from the compilations of Coward and Jones 16 and Zabetakis. x7 Some of the values shown may be systematically higher than the true limit flame temperatures because of the use of electric spark ignition sources in the traditional method for measuring limits. 16'17 In some cases spark sources are inadequate even for homogeneous mixtures and measured lean limits with stronger sources are systematically lower than for spark sources. 4'13 H y d r o g e n is not shown in the figure, but it has a uniquely low limit flame t e m p e r a t u r e o f only 700 K. T h e less reactive amines have somewhat higher limit flame t e m p e r a t u r e s than do the saturated hydrocarbons from which they are derived. T h e less reactive halogenated hydrocarbons have still higher limit flame temperatures. For the first few members of each homologous series, there are wide differences in limit flame t e m p e r a t u r e s because their reactivities vary widely; however, as the carbon numbers increase, the limit flame temperatures converge "toward the saturated alkane value because the reactivities o f the fuels approach those o f the saturated alkanes.
d I.-- 1 , 5 0 0
FIG. 3. --Calculated lean limit flame temperatures as a function of carbon number for the homologous series of various hydrocarbons and substituted hydrocarbons.
In the dust flame p r o p a g a t i o n process, the pyrolysis and devolatilization o f the coal particle precedes the combustion process, and it generates volatiles a n d a char residue. The volatiles mix volumetrically with the air, and their exothermic oxidation sustains the flame propagation process. 2 T h e char oxidation reactions, on the other hand, are constrained to occur on a two dimensional surface, and that limitation causes the char oxidation rate to be too slow to make a significant contribution to flame propagation. A volatility model is therefore required in o r d e r to proceed further with this analysis. Comprehensive studies o f the chemical corn-
position of pyrolysis volatiles generated from coal were r e p o r t e d by several investigators. 19'2~ Their data, obtained at high heating rates, were used to obtain the following composition for the pyrolysis products generated from Pittsburgh seam bituminous coal: Gases: 1.32 H2 + 5.90 H 2 0 + 0.78 H2S + 3.28 CO + 1.71 CO2 + 3.42 CH4 + 1.71 C2H4 + 2.76 C3H6. T a r Volatiles: 34.12 CHa.a O0.06 N0.0b Char: 40.45 [treated as inert]. Ash: 2.48 SiO2 + 1.24 A1203 + 0.83 FeO. The coefficients listed are the weight percent o f the various molecules and substances. In the thermodynamic calculations, the char is represented as a fictitious inert element which cannot enter into any chemical reactions, but whose other t h e r m o d y n a m i c properties are taken to be identical to t h o s e o f carbon. T h e air is taken as 4 moles o f Oz to 15 moles of N2 (21.05 pct oxygen). The m a x i m u m volatility ~(ields obtained from wire-mesh heating studies 2~ are systematically lower than the values measured by d r o p furnace methods 21 or by laser heating. 2 2 T h e Bureau of Mines m e t h o d 22 uses single coal particles or small arrays o f individually heated particles subjected to laser heating for a controlled exposure time and flux. Weight loss measurements were obtained directly, and the maximum volatility yield from Pittsburgh seam bituminous coal was 55 pct. That value was used in the above volatile tabulation. T h e resultant t h e r m o d y n a m i c calculations made with that volatility model are shown in Fig. 4. T h e calculated adiabatic flame tempera2,600
"'"~uas% l Experirne.,
/ ,/ .... .... ~,.:s ........ I /..'" \\ \'\ ........ I1.." \ \
I- 1,800 j~ 1,600 LL
I ,(300 0
COAL DUST CONC, g/m s
FIG. 4. --Calculated adiabatic flame temperatures for constant pressure and constant volume combustion of coal dust in air compared with measured explosion temperatures at constant volume.
tures are shown as a function o f coal dust concentration u n d e r the assumption that all the volatiles participate in the combustion reactions. The constant pressure (HP) and constant volume (UV) t e m p e r a t u r e s are shown. T h e adiabatic predictions are c o m p a r e d with the measurements of m a x i m u m explosion temperatures. T h e measurements were made with infrared methods in an 8-L chamber; the detailed results were r e p o r t e d elsewhere. 23 T h e infrared methods give both particle temperatures and gas temperatures, and it is the latter that should be c o m p a r e d with the calculated adiabatic explosion temperatures. Gas temperatures were m e a s u r e d to be systematically higher than particle temperatures, but because o f emissivity limitations, the gas temperatures could not be m e a s u r e d reliably for leaner mixtures. T h e m a x i m u m temperatures occur at coal dust concentrations that are near stoichiometric with respect to the volatiles, which is consistent with the model. An extrapolation of the measured gas explosion temperatures to those concentrations gives a measured value that is somewhat lower than the adiabatic, UV value, and that is to be expected since the real system is not adiabatic. T h e major difference between the m e a s u r e d t e m p e r a t u r e data and the theory occurs in the behavior o f the nominally rich dust concentrations. T h e theoretical curves, which assume that all the volatiles participate in the combustion reaction, predict a much more m a r k e d decline for rich mixtures than 'is measured. A l t h o u g h there is a small decline in the m e a s u r e d temperatures at concentrations above stoichiometric, those temperatures eventually level off and become nearly constant at high concentrations. T h e theory, which assumes complete devolatilization, predicts a r a p i d decline for such rich mixtures. T h e same discrepancy is observed if one compares the m e a s u r e d pressures 4'6'27from coal dust explosions in the 20-L chamber with the adiabatic predictions u n d e r the UV constraints. T h e comparison is shown in Fig. 5 and, again, the data show no evidence of rich limit behavior even though the t h e r m o d y n a m i c predictions give the expected decline in explosion pressures for rich mixtures. Those differences will be addressed shortly, and will be explained in terms o f devolatilization rate limitations; however, for lean concentrations o f coal dust where the burning velocities are low and the time available for devolatilization is long, the rate of devolatilization is not limiting, a n d the assumption that all of the volatiles participate in the combustion reaction is reasonable. Accordingly, the lean limit prediction is considered first. For the
VOLATILITY MODEL FOR COAL DUST FLAMES composition of the coal volatiles chosen, the choice of a limit flame temperature that is close to the value chosen for polyethylene is reasonable. That value of 1,380 K is shown as the horizontal arrow in Fig. 4, which intersects the HP temperature curve at a predicted lean limit concentration of 90 g/m 3. The measured value, reported elsewhere, 4'~'27is 90 _+ 10 g/m 3, in good agreement with the prediction. T h e data in Fig. 5 show that as one increases the dust concentration from its near limit values to concentrations that are near-stoichiometric with respect to the volatiles, the measured explosion pressures are systematically about 75 pct of the calculated adiabatic ones. That is to be expected since the actual explosions are not adiabatic for a variety of reasons. Buoyancy distortions are inevitable, causing the u p p e r hemisphere to contact the top of the chamber wall before combustion is completed in the lower hemisphere. 1~'24 Such buoyancyinduced distortions generate nonadiabaticities. Other nonadiabatic effects are caused by radiative losses from the developing fireball, ~5 the nonsphericity of the enclosure or the propagating flame front, and heat losses to internal structures such as the dust probes and ignition leads. Although the real explosion is nonadiabatic as expected, the measured explosion pressure curve does properly parallel the predicted one at concentrations above the lean limit until one approaches the rich mixtures. For the higher dust concentrations in Figs. 4 5, the calculations of both flame temperature and explosion pressure, which are based on a constant and complete volatility yield of 55 pct, predict marked declines that are not observed experimentally. Those differences between the thermodynamic expectations and the data are attribut-
able to limitations on the rate of devolatilization. The data show that explosion pressures (and temperatures) tend to retain their peak values as dust concentrations increase rather than to decline as predicted by a volatility model, which assumes a constant and complete yield of volatiles d u r i n g combustion. T h e p h e n o m e n o n has been explained qualitatively in terms of the flame front feeding on just enough volatiles to "ride the crest" of a near-stoichiometric concentration in the gas phase. 2'5 Although for nominally rich mixtures, more volatiles may be generated after the flame front goes by, they are emitted too late to dilute the flame zone with excess fuel vapor, hence the thermodynamically predicted reduction in explosion pressure is not observed. Such a devolatilization rate limitation becomes controlling not only at higher dust loadings as shown in Figs. 4 and 5, but for large particle sizes and at high b u r n i n g velocities. Such a devolatilization rate limitation not only explains the data in Figs. 4 and 5, but it also explains several other observations: the absence of a normal rich limit of flammability for dusts, 4 the particle size dependences for their lean limits and autoignition temperatures, 2 and their flammability behavior at reduced O2-1evels, where the normal "nose" of the flammability curve for homogeneous gas is transformed into a "blunt brow" for dusts. ~ For a more quantitative comparison, one determines the fraction of volatiles that must be assumed to contribute to the flame in order to obtain perfect agreement between the measured explosion pressures and the calculated ones after allowing for finite heat losses. The ratio of contributing volatiles to the total volatiles available in the coal is defined as the [3-factor, and the values inferred from the coal dust explosion data are shown in Fig. 6 as a
KEY Theory Experiment 2
COAL DUST CONCENTRATION, g/rn 3
CO.&L DUST CONC, g/m 3
Fie. 5. --Calculated adiabatic explosion pressures for the constant volume combustion of coal dust in air compared with measured explosion pressures in the 20-L chamber.
Fie. 6. --Fraction of coal volatiles, 13, assumed to contribute to flame propagation in order to obtain agreement between measured explosion pressures and calculated pressures for constant volume combustion.
function of the initial coal dust concentration. The values shown in Fig. 6 can be rationalized as follows. T h e flame front heating flux which drives the devolatilization wave at a finite velocity t h r o u g h each particle z'z2'26 is p r o p o r tional to the b u r n i n g velocity of the coal dust-air flame, S,. T h e time available for the generation o f volatiles that can contribute to the flame p r o p a g a t i o n process is limited to the travel time of the flame t h r o u g h one flame zone thickness, 8 = a/S,,, where a is the effective diffusivity. T h e time available for devolatilization is therefore 9 = 8/S,~ = ot/S~ 2. T h e product of the devolatilization wave velocity (which is proportional to S~) with T, gives the d e p t h o f penetration o f the devolatilization wave. T h a t depth of penetration thus varies inversely with S,, as was also indicated previously. 2 For lean concentrations near the flammability limit, S, is small and that d e p t h of penetration of the devolatilization wave exceeds the particle radius, which means that the particle devolatilizes completely within the flame front, which gives [3 = 1 for the lean mixtures. However, as the dust concentration increases, the burning velocity increases rapidly, 27'2s and the d e p t h of penetration, which varies inversely with S,, decreases until its value becomes less than the average particle radius. For a coal dust concentration of 250 g/m ~, where the b u r n i n g velocity approaches its m a x i m u m value, !3 = 0.70, which means that some 30 pct of the coal dust mass is unable to devolatilize in time to contribute to flame propagation. At still higher dust concentrations, the b u r n i n g velocity remains high, but a decreasing fraction o f the mass of coal dust can devolatilize in time to contribute. T h e combustion wave is then constrained to "feed" on the finer dust particles, or the sharp corners and surface regions of the coarser particles. The excess mass o f coal, that is, its undevolatilized fraction, plus the residual char at the higher dust loadings begin to absorb a larger fraction of the flame's heating flux. But since the excess coal makes no contribution to the propagation process, the effective heating flux that drives the fresh pyrolysis wave ahead of the flame front is diminished. Thus as the coal dust concentration increases still further, the pyrolysis wave regresses m o r e slowly into each particle, and there is a smaller contribution of volatiles from each particle. But that smaller contribution is compensated for by the higher initial concentration o f particles. As a result, the combustion wave continues to "ride the crest" of a near stoichiometric concentration of volatiles even t h o u g h the nominal fuel loading of dust is very rich. As the wave "rides the crest" of the near-stoichiometric concentration of vola-
tiles, it continues to generate high explosion pressures and t e m p e r a t u r e s even at high dust loadings. Such behavior is characteristic of all heterogeneous dust flames, and it distinguishes them markedly from homogeneous gas ones.
Addition o f Solid Inhibitors to Methane-air
A solid inhibitor dust dispersed in an otherwise homogeneous fuel-air mixture is also a singly heterogeneous system as was coal dust in air. T h e data for two such inhibitors, CaCO3 (rock dust) and NH4HzPO4 (ABC powder), tested against 10 pct methane in air are shown in Fig. 7. T h e surface-area-weighted mean diameter, D~, for the rock dust was 10 to 20 p~m. Two sizes of NH4HzPO4 were studied: regular ABC with Ds = 18 ~m, and ultrafine ABC that was centrifugally classified to D, = 3 p~m. For the data shown in Fig. 7, the inerting levels required to completely suppress stoichiometric methane-air explosions were measured with ignitors that were substantially m o r e energetic than those used in an earlier study? For rock dust, the data still show that it is virtually ineffective in suppressing the homogeneous gas explosion even at concentrations as high as 5,000 g/m 3. T h e data for the regular size ABC gives an inerting limit of about 1,300 g/m 3 with the more energetic 1,000 J ignitor. T h a t level is a factor of two higher than the 650 g/m 3 value obtained for r e g u l a r ABC with the less energetic matches? T h e inerting limit shown in Fig. 7 for the ultrafine ABC is 250 g/m 3, and that value is the same as the 250 g/m 3 r e p o r t e d earlier with the weaker source. T h e comparison between the measured and calculated explosion pressures for fluidized rock dust is shown in Fig. 8. For comparison, the experimental curve was "normalized" in order to account for its nonadiabaticity, by increasing the m e a s u r e d explosion pressures by
ZOO 4 0 0
1,000 1,200 1,400 ~ 2,000 INHIBITOR DUST CONC, g/rn 3
NH4N2P04, uBraf i ~ caco s
Pure OH4 3,000
FIe. 7. --Effect of added inhibitors on the peak explosion pressure of near stoichiometric methaneair mixtures.
VOLATILITY MODEL FOR COAL DUST FLAMES
According to this analysis, the only way one can be confident that a solid inhibitor is acting "chemically," is if its J3-factor exceeds unity. It is only when the measured inerting concentration is below that calculated on the basis of thermodynamic equilibrium that one can be confident that the inhibitor is acting "chemically" by reducing the reaction rates in the system, and thus preventing the exothermic combustion reactions from attaining their adiabatic equilibrium state.
KEY Theory, UV Experiment (normolized)
==5 3 4
INHLBITOR DUST CONC, g/m 3
Mixtures of Inhibitors and Coal Dust in Air FIG. 8. --Calculated adiabatic explosion pressures for the constant volume combustion of stoichiometric methane-air as a function of added rock dust (CaCO3) compared with the normalized, measured explosion pressures. about 16 pct, so that the measured value coincided with the adiabatic prediction at zero a d d e d dust concentration. T h e comparison shows that the actual effectiveness of the CaCO3 in reducing explosion pressures is much lower than would be expected on the basis of thermodynamic equilibrium. Clearly, the rock dust particles are not being heated uniformly to the final gas temperature. T h u s they are absorbing very little of the combustion energy. T h e rock dust particles have their individual heat transport and devolatilization rate limitations during the finite time of flame front passage. T h e adiabatic calculation indicates that a rock dust concentration of 450 g/m 3 should reduce the measured stoichiometric explosion pressure by 25 pct, yet the m e a s u r e d concentration required to achieve that reduction was 2,400 g/m s. Thus the effective [3-factor at 2,400 g/m 3 is only 0.2. At the higher concentrations, the effective B-factors are even lower. An HP calculation for p u r e CaCO3 a d d e d to stoichiometric methane-air indicates that a concentration of 490 g/m 3 would reduce the adiabatic flame t e m p e r a t u r e to its limit value of 1,450 K. T h a t 490 g/m 3 is therefore the calculated inerting level for CaCO3 addition u n d e r the assumption that all the dust reaches the final adiabatic equilibrium temperature. Yet the measured CaCO3 inerting level is in excess of 5,000 g/m 3, which corresponds to a i3-factor of less than 0.1. If a similar comparison is made for the regular NH4H2PO4 powder, its [3-factor at the inerting limit is 0.25. For the uhrafine NH4H2PO4, the measured inerting concentration is comparable to the calculated adiabatic value, and hence its effective ]3-factor is near unity.
For mixtures of coal dust and powdered inhibitors dispersed in air, the system is doubly heterogeneous, and devolatilization rate limitations can play a role for both the solid fuel and the solid inhibitor. T h e data for the inerting levels for CaCO3 a n d regular size NH4HzPO4 addition to coal dust are shown in Fig. 9. The data are expressed in percent by weight of the inhibitor in the coal-inhibitor mixture, and were obtained with 5,000 J ignitors. T h e initial addition of the NH4H2PO4 inhibitor raises the coal lean limit slightly, but as more inhibitor is added, the lean limit increases more rapidly until, at a coal dust concentration of 400 to 600 g/m 3, the full inerting level is finally attained. At that coal dust concentration, the measured inerting levels, when expressed in terms of the mass concentration o f inhibitor dust, correspond to a regular size ABC concentration of 320 to 500 g/m 3. For rock dust the measured inerting level corresponds to a concentration of 1,200 to 1,800 g/m 3. Theoretical calculations for the effect of addition of the inhibitors to a stoichiometric 600
_'~ 500 KEY NH4H2PO4
300 200 u. lOG
I L L I 1 20 30 40 50 60 INHIB['rOR IN MIXTURE, pr
FIG. 9. --Effect of added inhibitors on the flammability limit of Pittsburgh seam bituminous coal dust, indicating final inhibitor levels required to completely inert the dust mixture.
ratio of coal volatiles in air give predicted inerting concentrations of 420 g/m ~ for the NH4H2PO4 and 620 g/m 3 for CaCO3. T h e predicted ABC level is within the range of the measured values, a n d one could argue that it devolatilizes completely in the coal dust flame. For the rock dust, there is little doubt that its devolatilization rate is still limiting its effectiveness. The ratio of predicted-to-measured inerting concentrations corresponds to a [3-factor of 0.3 to 0.5. Direct measurements of the laser pyrolysis rate of CaCO3 show that its intrinsic devolatilization rate is about a factor of 4 slower than that of coal dust at high heating fluxes, 5 which is quite consistent with that measured [3-factor. T h e CaCO3, which was almost completely ineffective against stoichiometric methane-air, shows some moderate effectiveness against coal dust. The relaxation of the devolatilization rate requirements of the rock dust from an absolute one against the homogeneous, stoichiometric methane-air to a relative one against the heterogeneous coal dust results in a marked increase in its effectiveness. From a virtually ineffective additive with respect to methane ([3 < 0.1), it has a finite effectiveness against coal dust (13 = 0.3 to 0.5). I n m i n i n g situations, methane is a persistent problem, and hence there is a strong incentive for finding an inhibitor that is better than rock dust, one that would be effective against both coal dust and methane.
Conclusion The addition of a simple volatility model to the established concepts and mechanisms for homogeneous flame propagation is sufficient to explain the propagation and extinguishment behavior of heterogeneous coal dust explosions as well as other heterogeneous sytems containing solid inhibitors. For the finite time of flame front passage, the limitations on the rate of heating and/or devolatilizaton of the solid dust particles play a key role. Including those limitations in a volatility model is sufficient to explain the measured explosion pressures, flammability limits, and extinguishant requirements. Homogeneous gas flames are much more difficult to extinguish with powdered inhibitors than are heterogeneous dust flames because of the absence of such a devolatilization rate limitation when the fuel is initially present as a premixed gas. The data also show that such limitations play an important practical role in extinguishment. The particle sizes used in the commercially available extinguishant powders were designed for use in fires or diffusion flames
where residence times in the flame zone are long. But for the much shorter residence tintes involved in explosion extinguishment, their diameters are still too coarse for achieving optimum effectiveness, an effectiveness that would be u n e n c u m b e r e d by devolatilization rate limitations.
REFERENCES 1. HER'rZBERG, M., CASHDOLLAR, K. L. AND LAZZARA, C. P.: Eighteenth Sympositun (Imernational) on Combustion, p. 717, The C(mlbustion Institute, 1980. 2. HERTZBERG, M., CASHDOLLAR, K. L., NG, D. L. ANDCONTI, R. S.: Nineteenth Symposium (International) on Combustion, p. 1169, The Combustion Institute, 1982. 3. HERTZBERG, M., CONTI, R. S. AND CASHDOLLAR, K. L.: Twentieth Symposium (International) on Combustion, p. 1681, The Combustion Institute, 1984. 4. HERTZBERG, M., CASHDOLLAR, K. L. ANt) ZI.OCH-
OWER, I. A.: Flammability Linfit Measurements for Dusts and Gases; Ignition Energy Requirements and Pressure Dependences, This Symposium, 1986. 5. HERTZBERG,M., CASHDOLLAR,K. L., ZLOCHOWER, I. A. AND NG, D. L.: Twentieth Synlposium (International) on Combustion, p. 1691, The Combustion Institute, 1984. 6. CASHDOLLARK. L., SAPKO,M.J., WEISS,E. S. AND HERTZBERG, M.: Laboratory and Mine Dust Explosion Research at the Bureau of Mines. Presented at the Symposium on Industrial Dust Explosions, l 0-12 June 1986, Pittsburgh, PA, to be published by ASTM as STP 958. 7. U.S. Congress, Federal Mine Heakh and Safety Act of 1977. Public Law 95-164, Nov. 9, 1977. 8. American Society fbr Testing and Materials. Proo of the Symposium on Industrial Dust Explosions, held in Pittsburgh, PA, June 1986. To be published as STP 958. 9. BARTKNECHT, W.: Explosions, Course, Prevention, Protection; Springer Verlag, 1981. 10. BURGESS,M.J., ANDWHEVLER,R. V.:J. Chemical Soc. 99, 2013 (1911). 11. SIMMONSR. F., AND WOLFHARD, H. (;.: Combustion and Flame, 1, 155 (1957). 12. ZABETAKIS,M. G., LAMBIRIS,G. AND SCOTT, (;. S.: Seventh Symposium (International) on Combustion, p. 484, The ('ombustion Institute, 1958. 13. HERTZBERG, M.: In Fuel Air Explosions, Univ. Waterloo Press, p. 3, 1982. 14. GORnON,S., ANU McBRmE, B. J.: NASA SP-273 (1976). 15. JANAF. Thermochemical Tables, The Dow Chemical Co., Midland, MI.
VOLATILITY MODEL FOR COAL DUST FLAMES 16. COWARD H. F., AND JONES, G. W.: Bureau of Mines Bulletin 503, 1952. 17. ZABETAKIS, M.: Bureau of Mines Bulletin 627, 1965. 18. BURGESS,D. S., FURNO, A. L., KUCHTA,J. M. AND ML'RA, K. E.: Bureau of Mines RI 8709, 1982. 19. SOLOMON,P. R., HAMBLEN, D. G., CARENGELO,R. M. AND3K~USE, J. L.: Nineteenth Symposium (International) on Combustion, p. 1,139, The Combustion Institute, 1982. 20. SUUBERG,E. M., PETERS, N. A. ANn HOWARD,J. B.: Seventeenth Symposium (International) on Combustion, p. 117, The Combustion Institute, 1978. 21. KOBAYASHI, H., HOWARD, J. B. AND SAROFIM, A. F.: Sixteenth Symposium (International) on Combustion, p. 411, The Combustion Institute, 1976. 22. NG, D. L., AND I-IERTZBERG, M.: A Microscopic and Kinetic Study of Coal Particle Devolatiliza-
tion in a Laser Beam, to be published in NATO Advanced Research Workshop on "Fundamentals of Physical-Chemistry of Pulverized Coal Combustion" ed. J. Lahaye, Proc. of workshop held at Les Arcs, France, July 28-Aug. 1, 1986. CASHDOLLAR, K. L., aND HERTZBERG, M.: Combustion and Flame, 51, 23 (1983). SAPKO, M., FE'RNO, A. L. AND KUCHTA, J. M.: Bureau of Mines RI 8176, 1976. HERTZBERG, M., JoHNsor,', A. L., KUCHTA,J. M. AND FURNO, A. L.: Sixteenth Symposium (International) on Combustion, p. 767, The Combustion Institute, 1976. LEE, C. K., SINGER, J. M. AND CHAIKEN, R.: Combust. Sci. and Tech., 16, 205 (1977). CASrfD3OLLAR,K. L., AND HERTZBERG, M.: Review of Scientific Instruments, 56, 596 (1985). SMOOT, L. D., HORTON, M. D. AND WILLIAMS, G.: Sixteenth Symposium (International) on Combustion, p. 375, The Combustion Institute 1976.
COMMENTS P. Wolanski, Warsaw Univ. of Technology, Poland. What is the limit of application of the model regarding the volatile contents in dust particles, especially for the dusts with low volatile contents?
Authors' Reply. The flame propagation mechanism described here involves the devolatization of the dust followed by the homogeneous gas-phase combustion of the volatiles-air mixture. That mechanism adequately accounts for the explosion behaviour of almost all carbonaceous dusts in air. There may however be some exceptions. Fine graphite or even diamond dust are explosive at high dust concentrations at enriched oxygen levels. In those instances, an alternative mechanism involving heterogeneous surface oxidation may become plausible. A similar alternative mechanism may be required for some metal dusts that are explosive even though they have very high boiling points. The key parameter in determining whether the alternative heterogeneous surface oxidation mechanism is required, is the difference between the adiabatic limit flame temperature and the temperature at which the dust would display a significant volatility or vapor pressure. For carbonaceous dusts the limit flame temperature is high enough to ensure a plentiful yield of volatiles at the limit concentration. But that is not necessarily true for the other cases mentioned above. Clearly each type of dust has to be considered on a case by case basis.
P. Ronney, Princeton Univ., USA. The burning velocity at the flammability limit (which is dependent
on the experimental apparatus) is relatively constant among different families of fuels and thus would be a more useful predictor than a limiting flame temperature. If the effective overall activation energy of the fuel (which is a quantitative measure of the fuel's "reactivity" you mention) is known, the flame temperature, and thus the stoichiometry, necessary to produce the required limit burning velocity can be calculated by~any of a number of models of 1-D flame propagation based on large activation energy asymptotics. This analysis would provide a more widely applicable procedure for estimating flammability limits.
Authors' Reply. It is certainly true that the limit burning velocity for homogeneous mixtures is almost invariant among the different families of fuels. The reason for that constancy has been discussed in detail in reference (13) and in Bureau of Mines RI's 8127 (1976); 8469 (1980); 8607 (1982) and 8865 (1984). But that invariance does not necessarily mean that the burning velocity is a more "useful" predictor than the limit flame temperature, especially if one is dealing with a heterogeneous mixture of a dust in air. For such dust flames, the overall propagation mechanism is much more complex than for a homogeneous gas flame. Burning velocity predictions for such dust flames involve many more unknown kinetic parameters and are considerably more uncertain than is the calculated adiabatic flame temperature used here. One must, of course, be aware of the kinetic factors and the role they plan in the choice of the appropriate limit flame temperature. That question is discussed in the text.