Professional Documents
Culture Documents
Section2/Chapter 11
lN: DfNenno, P.M., et al., Editors, SFPE Handbook of Fire Protection Engineering,
2nd Edition, Chapter 11, Section 2, 2/171-179 pp
SMOLDERING
COMBUSTION
T J. Ohlemiller
2-171
2-172 FIRE DYNAMICS
rate of oxygen transport to the reaction zone. The control via tL = AL2 (1)
transport rate occurs because the heat evolved during smol-
der initiation raises the local temperature and thus the local where fL is the time for the smolder zone to penetrate the
reaction rate, until all of the neighboring oxygen is con- layer of thickness, L; andA is a constant that can, at present,
sumed. Subsequently, the reaction continues to consume only be determined by experimental measurement of at least
oxygen as fast as it reaches the reaction zone, yielding a very one layer thickness.
low oxygen level locally, which limits the reaction rate. This relation and Figure 2-11.1 imply that a 10-m fuel
The subsequent evolution of the smoldering zone away layer, such as might be encountered in a landfill or coal mine
from the initiation region is heavily influenced by oxygen tailing pile, would require more than four years for smolder
supply conditions. If initiation occurs deep within a layer of penetration. Such a deep layer is unlikely to be uniform in
fine particles (sawdust, coal dust), for example, it will slowly practice and the smolder front movement would be domi-
work its way to the nearest free surface at a rate dictated by nated by buoyant convective flow in regions of lesser flow
oxygen diffusion through the particle layer. (The more resistance. However, this does illustrate how very slow some
coarse and loosely packed the particles, the greater the in- smolder processes can be.
fluence of buoyant flow through the fuel leading to predom- A well-insulated reaction zone is a key factor in the
inant upward spread.) When the smolder zone reaches the existence of stable, self-sustaining smolder at such ex-
free surface region, it will spread more rapidly over this tremely low rates. The heat loss rate cannot exceed the heat
region in response to local convective and diffusive oxygen
supply conditions. As will be seen, when smolder spread
over the surface region of a fuel layer is forced by airflow, its 400
response also depends on heat transfer considerations.
In examining self-sustained smolder propagation and
its response to oxygen supply conditions, dimensionality
is important, It is necessary to distinguish one-dimensional
from multi-dimensional configurations. It is further nec-
200
essary to discern whether the smolder zone is spreading in
the same or opposite direction as the net movement of
oxygen.
100
I I I [ I I I I
B
A FLEXIBLE POLYURETHANE FOAM,7
PB = 0.04 @m3
1
B UNRETARDED CELLULOSIC INSULATION,5 /
0
1/
200-300#m PARTICLES,8 PB = 0.05-0.08 g/cm3
Fig.2-11.2.Smolder velocity vs airflow velocity into reaction zone for nearly one-dimensional reverse smolder. pn is bulk density.
generation rate. In this case, the same factor that is slowing in fact, acts like an insulator for the reaction zone, increas-
the oxygen SUpplY rate, and therefore the heat generation ing its stability. On the other hand, Dosanjh et al, 6 point
rate (i.e., the thick layer of wood particles over the reaction out that this mode of smolder propagation can achieve a
zone), is also slowing the heat loss rate. steady-state only if, as a minimum, the energy released is
In the previous example, the smolder propagation pro- sufficient to heat the incoming air supply; otherwise it will
cess is inherently unsteady because of the time-dependent extinguish.
oxygen supply process. If oxygen is instead continually sup- Figure Z-I 1.2 shows measured reverse smolder veloci-
plied by a forced convective flow through the fuel layer, ties for several types of fuel as a function of airflow velocity
nearly steady propagation occurs, Such a configuration is through the fuel bed. The bulk densities of the fuel bed are
encountered in some incinerators and coal burners but all low but typical for these types of materials. Note that the
rarely in a smoldering fire, This configuration has been ex- airflow velocity range is also quite low, although higher
amined experimentally and modeled, 718and is a relatively flows are sufficient to move the fuel particles in the bed (i.e.,
well-understood smolder mode with underlying mechanisms an upward flow higher than approximately 0.01 to 0.02 ro/s
qualitatively similar to the transient case just discussed. would fluidize the fuel bed).
In this mode of reverse smolder propagation, oxygen Despite the considerable variation in the chemical na-
surrounds the fuel particles as they are heated by the ad- ture of these fuels, the smolder velocity is always of order
vancing smolder reaction zone. Thermal degradation of 10-4 rnJs. For the same air supply rate, the smolder veloci-
some fuels in the presence of oxygen is exothermic. This is ties do not vary much more than a factor of 2. This is con-
particularly true of cellulosic materials and this heat can be sistent with the idea that the oxygen supply rate, not reac-
sufficient to drive the smolder wave without any char tion kinetics, dominates the propagation process. The
oxidation, 6 In flexible polyurethane foams, the presence of differences with fuel nature that do exist mainly appear to
oxygen during degradation plays another key role. Without reflect variations in available heat and effective thermal
oxygen many foams do not form any char, 7 although char conductivity.
oxidation is a necessary source of heat for these materials. In Only limited information is available on toxic gas pro-
the reverse smolder mode, the net oxidation rate and net heat duction from this mode of smoldering. The molar percentage
release rate are again directly proportional to the oxygen of carbon monoxide in the evolved, gases has been examined
supply rate; the smolder zone spreads to adjacent material for two of the fuels in Figure 2-11.2. For the flexible poly-
as fast as this generated heat can be conducted and radi- urethane foam, the carbon monoxide is 6 to 7 percent for an
ated to it, An increased oxygen supply rate causes a greater air velocity of 1.5 x 10 3 rnk the flow rate dependency was
rate of heat release and increased peak temperature in the not examined, g For the cellulosic insulation material, 10 the
reaction zone which, in turn, increases the heat transfer carbon monoxide mole fraction varies from about 10 to 22
rate to adjacent fuel, thus accelerating the smolder spread percent from the lowest to the highest air flow velocity in
rate. This sequence implies that the smolder reaction zone Figure 2-11.2. The mass flux of carbon monoxide from such
may well move through a layer of fuel without fully con- a smoldering process (grams of CO/m2 of smolder front/
suming the solid at any point. This unconsumed material, second) then is estimated as follows
2-174 FIRE DYNAMICS
+ mGs)
Yco(rnoir (2) smolder also allows for more complete combustion of the fuel.
These and other differences between the two smolder modes
or
can be explained in terms of the differing wave structures. 6
Frandsen14 investigated the downward propagation of
Ycotpairvair + (I +) APsvsl (3)
smoldering in horizontal layers of peat as a model fuel for the
complex duff layer found on the floor of a forest. No external
Here Yco is the mass fraction of carbon monoxide in the
flow was imposed. This is essentially a diffusion-driven for-
evolved product gases (approximately equal to the mole frac-
ward smolder process forced to be one-dimensional in this
tion); mair is the mass flux of air entering the smolder zone;
study; it normally is multidimensional in character. The
m Gs is the mass flux of gaseous material evolved from the
influence of both moisture and inorganic diluents on the
solid fuel; pair is the density of the air at the point where its
limits of smolder propagation was measured. At the ex-
velocity, vair, is measured; + is the initial void fraction of the
tremes, it was found that this cellulosic fuel will just smol-
fuel bed; Ap, is the change in density of the fuel bed (for
der when it contains 50 percent water by weight and no
reverse smolder, typically 65 to 95 percent of the original
inorganic diluents; when dry it will just smolder when the mix
mass is gasified); and us is the smolder front velocity.
contains 80 percent inorganic diluents. These results should
Limited information is also available on the aerosol
be roughly indicative of the limits for other cellulosic fuels in
emitted by a reverse smolder source; II this is pertinent to
the absence of a crossflow over the fuel layer.
detection of a smoldering fire. The source studied was es-
sentially identical to that used to obtain the data for curve B
in Figure 2-11.2; the fuel again was an unretarded cellulosic
insulation. The mass mean particle size of the aerosol was 2 Multi-Dimensional Smolder Spread
to 3 p,m;this is about 5 times larger than cigarette smoke and Factors such as ignition source geometry, fuel geometry,
50 to 200 times larger than the sooty particulate produced by and the strong influence of buoyant flow on oxygen supply
flaming combustion. This large size explains the relatively usually interact to assure that a smolder reaction zone has
poor sensitivity of ionization smoke detectors to realistic significant gradients of temperature and species in two or
smolder sources. The residual solid left in the smolder wave three dimensions. The number of possible configurations
and the original fuel both were found to be effective filters becomes virtually limitless. The practical configurations
for this aerosol; this helps explain the observation by that have been studied are few and they are usually two-
Palmer5 that smoldering in a thick layer of fuel was not dimensional; they do shed some light on most cases likely to
detectable until it neared the surface exposed to the ambi- be of interest.
ent atmosphere. Horizontal fhel layer: The configuration that has been
The rate of heat release for this mode of smolder can be
studied most extensively is two-dimensional smolder prop-
estimated from the total mass flux of products and their heat
agation in a uniform horizontal layer of particles or fibers.
content (gas temperature typically 670 to 970 K). The result
Ohlemiller15 examined the structure of the smolder zone in
is a few kW/m2 of smolder front. This translates to a few
a thick (0.18-m) horizontal layer of cellulosic insulation in
hundredths of a kW for a reverse smolder source 0.1 to 0.15
the absence of any forced airflow over the fuel layer. In these
m in diameter. The strength of the heat source has a bearing
conditions, the flow induced by the buoyant plume rising
on the behavior of the buoyant plume. (See Section 2, Chap-
above the smolder zone assures a constant supply of oxygen
ter 2.) Sources as weak as those considered here generate
to the space above the layer; oxygen penetrates the layer
plumes that may not reach the ceiling of a room. 12
largely by diffusion.
Forward propagation: The second limiting case of one- If such a layer is ignited uniformly on one end, the
dimensional smolder propagation is called forward smolder; smolder reaction zone soon evolves into a new shape dic-
in this case the oxygen flow is in the same direction as the tated by oxygen supply rates. 15 The uppermost elements of
movement of the smolder front. The most familiar example the reaction zone, being closest to the free surface and hence,
(though not one-dimensional) of forward propagation is a ambient air, spread away from the ignition source the fast-
cigarette during a draw. This limiting case is encountered in est; successively deeper elements spread in the same direc-
some industrial combustion processes but is unlikely to be tion but more slowly. The result is a smolder reaction zone
found in its pure, one-dimensional form in a fire context that (viewed in vertical cross-section) slopes upward from
(some elements of this mode are encountered in realistic the bottom of the layer to the top, in the direction of move-
cases, however]. An approximate model of this process (in ment, The steady-state length of this inclined smolder front
one dimension) has been presented by Dosanjh and Pagni. 13 is roughly twice the depth of the original fuel layer. This
They point out that this smolder mode will die out if the heat inclined reaction zone is several centimeters thick, and
generated by char oxidation is insufficient to drive the dry- across this thickness there is a smooth transition from un-
ing and fuel pyrolysis reactions that precede char formation burned fuel to ash. On the ash side (the free surface adjacent
in the reaction zone. to air) oxygen diffuses down and inward in the same direc-
Some characteristics of forward propagation are briefly tion as the smolder front is moving and attacks the charred
mentioned here to describe the major effects that reversing fuel; this is analogous to forward smolder discussed earlier.
the direction of oxygen flow can have on smolder propaga- On the unburned fuel side of the inclined smolder front,
tion characteristics. oxygen diffuses in from the region ahead of the front to react
Forward and reverse smolder propagation have been with the fuel as it is thermally degraded by heat conducted
compared experimentally ;610 the fuel was an umet~ded from the char oxidation region. Oxygen here is moving OP-
cellulosic insulation. Forward smolder through this same posite to the direction of smolder propagation, so this aspect
fuel at the same air supply rate is about ten times slower than of the overall reaction zone is analogous to reverse smolder.
reverse smolder. The carbon monoxide mole fraction is in- Remember that in cellulosic materials, this oxidative/ther-
dependent of air supply rate and is about 9 percent. Forward mal degradation is exothermic. Thus the two-dimensiond
SMOLDERING COMBUSTION 2-175
lo- , , I gion, which receives the best oxygen supply, is also sub-
< SMOLDERING jected to the highest heat losses.
AND AIR The influence of two factors, fuel particle size and rel-
7- a FLOW IN
ative direction of airflow and smolder propagation is shown
/A SAME
DIRECTION
in Figure 2-11.3. Particle size has a relatively weak effect on
5t
//:// -
4 smolder velocity but its effect depends on whether the smol-
t A SMOLDERING AND i
AIR FLOW IN
der configuration is forward or reverse. The configuration
itself has a much greater effect.
3- / 0hlemiller17 obtained comparable smolder velocities
OPPOSING
and dependence on configuration for 0.10- to 0.1 l-m thick
.~ . A-&-- DIRECTIONS
layers of cellulosic insulation. It was found that the config-
y_ J uration dependence cannot be explained solely on the basis
of oxygen supply rates, The mass transfer rate to the surface
of the fuel bed was measured for forward and reverse con-
figurations; it differs by only 20 to 30 percent (these differ-
11 , 1 1 1 ! , 1 1 1 ences are caused by changes in the bed shape due to shrink-
40 60 80100 200 400 age during smolder). It was pointed out that the observed
dependence on relative direction of the airflow is consistent
Fig. 2-11.3. Dependence of smolder propagation rate through with there being a prominent role for convective heat trans-
horizontal layers of beech sawdust on air velocity over top of fer along the top surface of the fuel layer. This dependence
layer.4 Circles: 120 p.m mean particle size; triangles: 190 w occurs only if part of the smolder wave, i.e., the region near
mean particIe size; squares: 480 P mean particie size. the leading edge, is kinetically limited (and therefore highly
temperature sensitive) rather than oxygen supply rate lim-
ited. This explains the qualitative impact of both relative
horizontal smolder zone incorporates features of both for- airflow direction and combustion retardants on smolder ve-
ward and reverse smolder and is driven forward by the com- locity; it also explains why forward smolder is faster than
bined heat release from char oxidation and oxidative/ther- reverse smolder in the horizontal layer configuration,
mal degradation. whereas the opposite was true for one-dimensional propa-
The participation of oxidativehhermal degradation in gation. The role played by fuel particle size maybe implicit
driving the smolder process requires that oxygen have free in this view, but a quantitative model is not yet available.
access to the thermal degradation region. For a low-permea- In contrast to the monotonic enhancement of forward
bility fuel such as solid wood, this is not the case. Even smolder velocity with increased airflow rate found by
though solid wood has basically the same reaction chemistry Palmer and by Ohlemiller, Sato and Sega18 observed more
as cellulosic insulation (which consists mostly of wood fi- complex behavior with thin [0.004 to O.ol-m) layers of a
bers) and smolders with a qualitatively similar inclined re- cellulosic mixture. Smolder velocity increased up to
action zone, it must be driven solely by char oxidation. freestream air velocities of about 3 n-ds and then remained
The low permeability and corresponding high density constant to the highest air velocity examined (6 m/s]. This
of solid wood has another consequence with regard to smol- plateau correlated with erratic behavior at the leading edge
der. The self-insulating quality of the reaction zone is much of the smoIder reaction zone involving both periodic extinc-
less than with a low-density layer of fuel particles or fibers. tions and mechanical disruptions. These authors also exam-
A single layer of wood will not sustain smolder unless it is ined the thermal structure of their forced smolder waves.
subjected to an additional heat input of about 10 kW/m2;16 The results were qualitatively similar to those of Ohlemiller
this heat could come from some external radiant source or for buoyant smolder, 15 but the peak temperatures were ap-
from another piece of smoldering wood that has an adequate preciably higher due to the enhanced oxygen supply rates.
radiative view factor with respect to the first piece. There is a minimum thickness below which a horizon-
In view of the strong role of oxygen supply rate in shap- tal fuel layer will not undergo self-sustained smolder prop-
ing the smolder process in a horizontal fuel layer, it is not agation. As the thickness of a fuel layer decreases, its
surprising that smolder also accelerates in response to an surface-t o-volume ratio increases (inversely with thickness
increased oxygen supply rate produced by an airflow over to the first power). The ratio of the rate of heat loss to the rate
the top of the smoldering layer. As with the one-dimensional of heat generation also varies in this manner so that ulti-
propagation situation, two possibilities again exist: the air- mately the losses are overwhelming and extinction occurs.
flow can travel in the same direction as the smolder front The exact thickness will depend on factors such as bulk
(again called forward smolder] or in the opposite direction density, fuel type and particle size, rate of oxygen supply,
(reverse smolder). Note, however, that now the actual fluxes etc., influencing the heat generation per unit volume at a
of oxygen witl+n the smoldering fuel bed may go in various given thickness. The same considerations apply to other
directions; they are no longer constrained to being parallel to thin layers of fuel such as fabrics on upholstery and sheets of
the smolder wave movement, adin the one-dimensional cases. paper, wood, or particle board. Pahner5 found that the mini-
Palmer5 examined both of the flow direction possibili- mum depth for sustained smolder instill air increased linearly
ties for relatively thin horizontal layers (3 x 10-3 to 5.7 x with particle size for beech, pine and cork; for cork this depen-
10-2 m) of various cellulosic particles (cork, pine, beech, dence ceased above 2 mm, apparently because more complete
grass). Figure 2-11.3 shows some typical results. Note that oxidation of the char stablized the process in the layers of
the smolder velocities are less than or equal to those in larger particles. For very small particles, (< 100 pm), the
Figure 2-I 1.2, despite the much higher air velocities. This is minimum depth dropped as low as 1 mm for cork dust; 0.01 m
probably due to differing rates of actual oxygen delivery to was typical of smell particles of beech or pine sawdust.
the reaction zone, and to the fact that the near-surface re- Ohlemiller and Rogers 19 found the minimum depth in still
2-176 FIRE DYNAMICS
air for an unretarded cellulosic insulation to be O.035 m; a content yielded smolder spread, 10 percent fuel content did
heavy loading of the smolder retardant boric acid roughly not. The depth to which this spreading smolder zone
doubled this value. Since the insulation has a very small reached increased with the cross-sectional dimensions of
effective particle size and essentially the same chemistry as the fuel bed. Material below this depth, having been heated
Palmers sawdusts, most of the difference in minimum depth and partially decomposed by the smoldering zone above,
(for the unretarded material) probably lies in the bulk den- could itself subsequently propagate a second wave moving
sity, which is about four to five times less for the insulation in the opposite direction. It was pointed out that such be-
compared to the sawdusts (4o kg/m3vs 180 kg/m3).Palmer havior can make it possible for a landfill or similar fire to
found that the minimum depth dropped rapidly with in- spread under a barrier intended to stop it. Similarly, inert
creased airflow over the sawdust layers, in keeping with the covering layers may simply slow but not stop such fires. The
idea that a greater rate of heat release per unit volume stabi- true key to stopping a smoldering fire is getting the heat out
lizes the smolder process. of the fuel, but this can prove to be extraordinarily difficult.
Beever20,21 has addressed a problem at the opposite Smolder propagation data on a few other fuels (includ-
extreme of layer thickness, that of underground fires in land ing some that are inorganic) in horizontal layers can be
fills, peat deposits, and mine tailings. These tend to be smol- found. g Unfortunately, no data are currently available on
dering fires in roughly horizontal layers where the principal the evolved products of horizontal layer smolder. For crude
mode of oxygen access is from the top surface. Beever21 estimates on cellulosic materials the previous results for
studied this process on a laboratory scale using mixtures of reverse smolder are adequate, but they should be applied
fine sawdust or charcoal with an inert diluent, i.e., diatoma- here with caution.
ceous earth, in a trough that was insulated on the sides and
bottom but open to quiescent air on the top. The trough was Other fuel configurations: Data on a few other multidimen-
0.13 m by 0.38 m in cross section and 0.14 m deep; a deeper sional smolder configurations are summarized in Table 2-11.1.
trough was used in separate experiments in which pure Again there is little more information available than the rate of
layers of the inert diluent were placed atop the combustible smolder propagation. An exception to this is the smoldering
layer. Local ignition near the top of a layer yielded steady cigarette, which has been extensively studied,22,2324 albeit
propagation over a limited depth at rates that varied only usually in a manner most pertinent to its peculiw mode of
weakly with inert content. However, while 25 percent fuel cyclicly forced air supply.
SMOLDERING COMBUSTION 2-177
Al the materials in Table z-11. I are fairly porous. As with experiment though some of the input parameters must
noted previously, solid wood, a low-porosity fuel, also smol- be forced slightly.
ders, given a configuration that limits heat losses. 25 In contrast to the above result, a more recent study of the
Ohlemiller and Shaub26 and Ohlemiller27 examined fire risks associated with upholstered furniture implied that
smolder spread along the interior surface of a three-sided the toxic exposure from a smoldering chair in an average
channel constructed of either white pine or red oak. A con- house was rarely fatal; transition to flaming brought with it
~ollecl flow of air was introduced at one end of the channel; death due to thermal causes.40 The methodology was indi-
the products evolved from the other end were monitored as rect; it involved using the Hazard I smoke movement and
was the rate of smolder spread. For both types of wood, tenability models in a reasonably successful effort to repro-
stable srno?der was observed for only a narrow range of inlet duce national fire statistics for upholstery fires. There are
ah velocltles, O.O5to 0.20 m/s. (From limited data this ap- not as yet sufficient data on the toxicity hazards of smolder-
peared true for both forward and reverse smolder.] Below ing upholstery materials to definitively resolve this issue.
this range the smolder process extinguished and above it flam-
ing eventually erupted. Both of these limits, but particularly
the lower limit, we probably dependent on the specific condi-
Transition to Flaming
tions of the tests. Cwbon monoxide typically comprised 2 to 3 The transition process from smolder to flaming in the
percent of the gases leaving the channel or about 10 to 15 above bedding and upholstery fires is essentially spontane-
percent of the gases leaving the surface of the wood. The rate of ous. At room conditions both smoldering and flaming are
heat release during smoldering was estimated from the oxygen possible in many such systems. Sato and Sega41 explored
consumption rate, correcting for carbon monoxide. This the domain of overlapping smolder and flaming potential for
ranged from about 0.5 to 2 kW or roughly 10 to 30 kW/m2, cellulosic materials and noted a hysteresis in the spontane-
based on the approximate area visibly glowing. ous transition between these two combustion modes. The
The last type of smolder configuration referenced in mechanism of such a spontaneous transition has not been
Table 2-11.1 is quite pertinent to the scenario that makes investigated in detail. It has been suggested on the basis of
smoldering a major contributor to residential fire deaths, i.e., small mock-up studies that a chimneylike effect develops in
upholstery and bedding fires initiated by cigarettes. This is the crevice between the horizontal and vertical cushions of
frequently a composite problem, with the smoldering ten- a smoldering chai~42 the enhanced air supply presumably
dency of both the fabric and the substrate (polyurethane accelerates local char oxidation, heating the char to the
foam, cotton batting) pertinent to the overall smolder behav- point where it can ignite pyrolysis gases. Such a mechanism
ior of the combined assembly. 28 Ortiz-Molina et al have is plausible but it has not been demonstrated to be operable
shown that the combination of a cellulosic fabric plus a in real upholstery or bedding, where the chimney effect may
polyurethane foam can smolder over a substantially wider not develop so readily,
range of conditions than can the foam alone. 29 The fabric Transition to flaming (fast exothermic gas-phase reac-
smolder process supplies added heat to the foam smolder tions) requires both a mixture of gases and air that are within
zone while simultaneously competing for oxygen. The full their flammability limits and a sufficient heat source to ig-
complexity of this interaction is yet to be explored, A con- nite this mixture. Furthermore, these two requirements
siderable amount of empirical data on the tendenc of ciga- must be realized at the same locus in space and at the same
rettes to initiate this type of smolder is available. 3IK36 time. Any factor that either enhances the net rate of heat
The life hazard posed by smoldering bedding or uphol- generation or decreases the net rate of heat loss will move the
stery within a closed room has been studied to some smoldering material toward flaming ignition by increasing
extent.37-39 Data have been presented 38 on the buildup of both local temperature and rate of pyrolysis gas generation.
carbon monoxide (near the ceiling) in a 2.4 m room on a side Such factors include an enhanced oxygen supply, an in-
due to cigarette-initiated smolder in a cotton mattress. The crease in scale (which usually implies lesser surface heat
smolder front was reported to spread radiaIly at a rate of 6.3 losses per unit volume of smoldering material), or an in-
10-5 m/sindependent of the size of the smoldering area. In creasingly concave smolder front geometry, which re-
two out of five tests the smolder process underwent a tran- duces radiative losses to the surroundings and enhances
sition to flaming combustion after 65 to 80 minutes, which is gaseous fuel concentration buildup. All of these factors may
close to the time at which total carbon monoxide ex osure be operating simultaneously in the case of upholstery and
was estimated to be lethal. Similar data are reported J 8 for a bedding smolder; sequential photos of smolder initiation,
greater variety of bedding and upholstery materials; these growth, and transition to flaming in an upholstered chair
were ignited by cigarettes (and by flaming sources) in a room appear consistent with this idea.42
4.3 x 3.6 x 2.4 m. Carbon monoxide and several other gases A further factor in this and in other systems involving
were sampled at three locations. Flaming developed from cellulosic materials is secondary char oxidation. This pro-
smoldering in several of the tests; this usually required 2 to cess is quite similar to the afterglow seen in cellulosic chars
3 hours of smoldering first. Again, the total exposure to left by flaming combustion. Intense, high-temperature
carbon monoxide from the smolder smoke approached or (probably greater than 107o K] reaction fronts propagate
exceeded lethal levels. Lethal conditions due to carbon mon- intermittently in seemingly random directions through the
oxide were reached in much shorter times in some cases. fibrous low-density char left by the main lower temperature
All available data on the hazards of smoldering in a smolder front. In charred fabrics, these #owin~ fronts can
closed room were evaluated;3g it was concluded that the sometimes progress in a stable manner along the charred
probability of a lethal carbon monoxide dose and of transi- residue of a single fiber, despite very high heat losses per
tion to flaming are comparable for a period from 1 to 2VZ unit volume of fuel. Such a process requires the catalytic
hours after cigarette initiation of smoldering. A model is action of alkali metals that are frequently found naturally in
presented for buildup of carbon monoxide due to a smolder- cellulosics or left there during manufacture .4344While in a
ing fire;sg the results generally show reasonable agreement very hot smolder front the size of a single fiber is unIikeIy to
.
be sufficiently energetic to ignite flammable gases, the larger rate has on the smolder process. The other very important
fronts (10 -3 to 10-2 m in scale) may well be. An analogous factor is the relative direction of movement of oxygen supply
process has been found to cause occasional flaming ignition and smolder propagation; this can be somewhat obscure in
of smoldering, unretarded celhxlosic insulation. 15 many realistic configurations. The actual chemical nature of
The transition from smolder to flaming can also be the fuel is relatively secondary, at least with regard to smol-
induced, for example, by a forced increase in oxygen sup- der rate. It maybe important for toxic gas production rates,
ply rate to the smolder reaction zone. 5194*45146This was but the data here are quite limited.
first studied quantitatively by Palmer 5 for airflow over
horizontal layers of wood sawdust; this process, of course, REFERENCES CITED
is familiar to anyone who has started a camp fire from
tinder and sparks. Transition to flaming was noted by 1. F. Clarkeand J. Ottoson,Fire J., 20,May (1976).
Palmer only for airflow in the same direction as smolder 2. R. McCarter,J. Cons. Prod. Flamm., 4, 346 (1977).
propagation [forward smolder); depending on the mate- 3. P. Bowes, SeIf-Heating: Evacuating and Controlling the Haz-
rial, the transition occurred at airflow velocities from ards, Chap.7, Elsevier,NewYork/Amsterdam (1984).
about 0.9 to 1.7 m/s. For these materials, flaming did not 4. T. Ohlemiller,Prog. in Energy and Comb. Sci., 11,277 (1985).
5. K. Palmer,Comb. and Fkme, 1, 129 (1957).
develop when the mean particle size was less than 1 mm. 6. T. Ohlemillerand D. Lucca, Comb. and Flame, 54,131 (1983).
0hlemiller45 did obtain transition to flaming in layers of 7. T. Ohlemiller,J. Bellan, and F. Rogers, Comb. and Flame, 36,
fibrous insulation materials of very small diameter ( 25 197 (1979).
pm) but again only with forward smolder; this occurred at 8. S. Dosanih,P. Pami, andC.Fernandez-Pello,Comb. and Flame,
air velocities of about 2 nds for unretarded insulation. 68,131 (1987).-
Leisch46 utilized ignition sources placed midway along 9. F. Roeersand T. Ohlemiller.1.Fire Flamm.. 11.32 [19801.
the length of grain and wood particle fuel layers so that 10. D. Lu~ca,An Investigation o~ Co-Current and Goun;er-C&mt
forward and reverse smolder zones were simultaneously Smoldering Combustion in Particulate Fuel Beds, MSE Thesis,
obtained; flaming was noted at 4 m/sair velocity only after Princetan(1979).
11. G.MulhoJlendend T. Ohlerniller,Aem Sci. and Tech., 1,59 (1982).
the smoldering process produced a substantial depression 12. H. Hotta, Y. Oka, and O. Sugawa, Fire Sci. and Technol., 7, 17
or cavity in the surface of the fuel layer. (1987).
0hlemiller45 explained the weak response and lack of 13. S. Dosanjh, and P. Pagni, Proc. of the 1987 ASME/JSME 7%ermal
flaming transition in reverse smolder on the basis of heat Engineering Joint ConferenceVolume 1, (P. Marto and L Ta-
transfer effects influencing the leading edge of the smolder nasawa, eds.) Book No. 10Z 19A, American Society of Mechan-
reaction zone. These heat transfer effects intensify the smol- ical Engineers, New York, (1987).
der in the leading edge region for forward smolder. In the 14. W. Frandsen, Can. J. For, Res., 17, 1540, (1987).
case of cellulosic insulation, the intensification leads to ran- 15. T. Ohlemiller, Comb. and Hame, 81,341 (1990).
16. T. Ohlemiller, unpublished test results.
dom development of small (a few cm) cavities near the lead-
17. T. Ohlemiller, Comb. and Flame, 81, 354 (1990].
ing edge which act as flame initiation regions and flame 18. K. Sato and S. Sega, Fire Safety ScienceProceedings of the
holders. Second International Symposium, Hemisphere Publishing
0hlemiller45 also found that both boric acid (a smolder Corp., New York, p. 87 (1989).
retardant) and borax (a flame retardant) could each elimi- 19. T.OhIerniller and F. Rogers, Comb. Sci. and Tech., 24139 (1980).
nate the transition to flaming when the retarded cellulosic 20. P, Beever, Subterranean Fires in the UK-the Problem, Build-
insulation was the only fuel. However, the effectiveness of ing Research Establishment Paper 1P 3/89, March, 1989.
the acid and borax was substantially reduced if the smolder- 21. P. Beever, Initiation and Propagation of Shouldering Reac-
ing fuel abutted unretarded wood; heat transferred from the tions, Ph. D. Thesis, Dept. of Physical Chemistry, Universityof
Leeds,June, 1986.
smolder zone readily ignited the wood. Palmer47 noted sim-
22. R. Baker, Nature, 247,405 (1974).
ilarly that layers of fine dust that would not themselves 23. R. Baker and K. Kilburn, Beitrage zur Tabakforschung, 7, 79
undergo transition to flaming readily ignited adjacent flam- (1973).
mable materials. 24. R. Baker, ibid, 11, 1 (1981).
Smoldering solid wood undergoes a transition to flaming 25. T. Ohlemiller and F. Rogers, Engineering Report No. 1432,
readily in a configuration that minimizes heat losses. 627 Princeton University, Princeton [1979).
It was inferred that the limiting variable in the transition is 26. T. Ohlemiller and W. Shaub, Products of Wood Smolder and
the surface temperature of the smoldering wood, with the Their Relation to Wood-Burning Stoves, National Bureau of
transition occurring when that temperature reached about Standards NBSIR 88-3767, May, 1988.
27. T. Ohlemiller, Fire Safety ScienceProceedings of the Third
95o to 1000 K.
International Symposium, Elsevier Science Publishing Co. Inc.,
New York, p. 565 (1991).
CONCLUSION 28. G. Tesoro, and T-Y. Toong, Smoldering in Cotton Upholsteqv
Fabrics and Fabric/Cushioning Assemblies, Mass. Institute of
Smoldering is a branch of solid fuel combustion quite Technology, Cambridge (1981).
distinct in many aspects from flaming, but equally diverse 29. M. Ortiz-Molina, T-Y. Toong, N. Moussa, and G. Tesoro, 17th
and complex. Unfortunately it has not been studied nearly to Symp. (Int.), Combustion Institute, Pittsburgh (1979).
the same extent as flaming. This is quite apparent in the lack 30. G. Damant, J. Cons. Prod. Flamm., 2, March (1975].
of quantitative guidelines that can be provided here for es- 31. G. Damant, J. Cons. Prod. Flamm., 2, 140 (1975).
32. G. Damant, J. Cons. Prod. FIamm., 6, 95 June (1979].
timating the behavior of realistic smolder propagation pro-
33. K. Palmer and W. Taylor, J. Cons. Prod. Flamm., 1, 186 [1974).
cesses, smolder detection, toxic gas production, and the 34. 1.Loftus.NBSIR78-1438, National Bureau of Standards, Wash-
transition to flaming. The experimental data provided can be ington (i978).
readily used for closely analogous situations; they must be 35. J. Krasny, Cigarette Ignition of Soft FurnishingsA Literature
used cautiously for dissimilar conditions. The reader should Review with Commentary, National Bureau of Standards
always bear in mind the strong role that the oxygen supply NBSIR 86-3509, Oct. 1986.
SMOLDERING COMBUSTION 2-179
36. R. Gann, R. Harris, J. Krasny, R. Levine, H. Mitler, and T. 41. K. Sato and S, Sega, ). Fire Sci., 3,26 (1985).
Ohlemiller,The Effectof CigaretteCharacteristicson the Igni- 42. R. SaIig, Smoldering Behatior of Upholstered Polyurethane
tion of Soft Furnishings,NationalBureauof StandardsTech- Cushioning and Its Relevance to Home Furnishings Fires, Mas-
nical Note1241, Jan. 1988. ters Thesis, Dept. of Mech. Eng., MIT, Cambridge (1 981),
37. K. Sumi and G. Williams-Leir, Research Paper No. 402,Na- 43. A. Ihrig, A. Rhyne, V. Norman, and A. Spears, j. Fire Sci., 4,237
tional Research Council, Ottawa (1969). (1986].
38. C. Hafer and C. Yuill, Characterization of Bedding and Uphol- 44. R. McCarter, J. Cons. Prod. Flamm., 4,346 (1977).
stery Fires, Southwest Research Institute, San Antonio (I 97o). 45. T. Ohlemiller, NBSIR 85-3222, National Bureau of Standards,
39. J. Quintiere, M. Birky, F. McDonald, and G. Smith, Fire and Washington (1985).
Mat.k, 6,99 (1982]. 46. S. Leisch, Smoldering Combustion in Horizontal Dust Layers,
40. W. Stiefel, R. Bukowski, J. Hall, and F. Clarke, Fire Risk As- Ph.D. Thesis, Dept. of Aero Eng., U of Mich., Ann Arbor, Mich-
sessment Method: Case Study 1, Upholstered Furniture in Res- igan (1983).
idences, National Institute of Standards and Technology 47. K. Palmer, Dust Explosions and Fires, Chapman and Hall, Lon-
NISTIR 90-4243, Mar. 1990. don (1973).