You are on page 1of 31

Report of Investigations 8469

The Theory of Flammability Limits

Conductive-Convective Wall Losses

and Thermal Quenching

By Martin Hertzberg

UNITED STATES DEPARTMENT OF THE INTERIOR

Cecil D. Andrus, Secretary

BUREAU OF MINES

Generated on 2015-01-31 00:24 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Lindsay D. Norman, Acting Director

This publication has been cataloged as follows:

Hertzberg, Martin

The theory of flammability limits. Conductive-convective

wall losses and thermal quenching.

(Report of investigations - Bureau of Mines ; 8469)

Bibliography p. 24-25-

Supt. of Docs, no.: I 28.23:8469-

1. Flame. 2- Combustion. I. Title. II. Series: United States. Bu-

reau of Mines. Report of investigations ; 8469.

Generated on 2015-01-31 00:25 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

TN23.U43 [QD516] 622s [628.9'22] 79-607931

SYMBOLS AND NOMENCLATURE

A The wall-loss surface area of a tube.

A The cross sectional area of a particle.

c The spatially averaged or effective heat capacity per gram.

c The heat capacity per gram of solid particle.

Cm The mass concentration of dust in a dust-air mixture.

Cm (max) Concentration of a coal dust at which its maximum S occurs.

d0 The diameter of a tube.

d The wall-loss quenching diameter or quenching distance.

d The diameter of a spherical dust particle.

ds The surface-area-weighted average particle diameter.

g The gravitational acceleration.

I The laminar heat flux on a particle.

Si The length of a tube.

L The kinetic theory mean free path for molecular collisions.

Nu The Nusselt number.

Pe The critical Peclet number for wall-loss quenching.

r The radius of a spherically expanding flame.

r0 The radius of a tube.

rp The radius of a spherical particle.

(Su)idea, The true or ideal laminar burning velocity.

Su The burning velocity. The velocity of a flame front relative to

the unburned gas.

(Su)a The horizontal limit burning velocity for quenching by natural

convection.

(Su)8 ef The upward limit burning velocity for quenching by flame stretch

Generated on 2015-01-31 00:25 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

(process e) induced by buoyancy (process a).

ii

(Su)a^ The downward stagnation limit burning velocity.

(Su)b The limit burning velocity for conductive-convective wall-loss

quenching.

(Su)'b The limit burning velocity for conductive-convective,

"internal-wall"-loss quenching by inert powders.

su (max) Tne maximum burning velocity for a coal dust-air flame which occurs

at a concentration c . .

m ( ma x)

t The time.

T The temperature.

T The temperature of burned gases.

Tu The temperature of unburned gases.

u The relative velocity of the gas with respect to the particle.

v The gas velocity.

vp The particle velocity.

V The volume of a tube.

V The volume of an individual particle.

x The propagation direction.

Ax The flame zone thickness.

a The spatially average or effective diffusivity.

3 The wall-loss factor (>1) which equals the number of flame-zone

thicknesses behind the flame front from which conductive losses

to cold walls can significantly influence the propagation process

at the flame zone.

6 The boundary layer thickness for radial heat conduction losses to

a cold tube wall.

n The gas viscosity.

X The spatially averaged or effective thermal conductivity.

v The kinematic viscosity of the gas.

p The spatially averaged or effective density of reacting, combusti-

ble, flame gas.

p The density of the solid comprising the dust particle.

Generated on 2015-01-31 00:25 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

T The characteristic heating time for a particle in the flame front.

T The characteristic heating time for the gas in the flame front.

63S

ill

CONTENTS

Symbols and nomenclature i

Abstract 1

Introduction 1

The conductive-convective wall-loss limit: quenching diameters 3

Buoyancy and flame shape for near limit flame propagation 6

The influence of apparatus dimensions and boundary constraints on the

measured limits of flammability 8

Thermal quenching by inert powders 14

Finite heat capacity effects; particle lag in the flame front; the

particle boundary layer 17

Conclusions 23

References 24

ILLUSTRATIONS

1. Source and sink power densities for the wall-loss quenching of a

flame propagating in a tube 3

2. Flow trajectories in combustion and buoyancy force fields 7

3. Flame shape and flame-zone structure for horizontal propagation in a tube 8

4. Lean limit measurements in methane-air mixtures for the three

directions of flame propagation in tubes of varying diameter,

and the measured burning velocities at those limit concentrations. 9

5. The sum of limit burning velocities for (a) buoyant-convective

quenching, and (b) wall-loss quenching, as a function of tube

diameter for the three directions of flame propagation 12

TABLES

1. Comparison of the sum of limit velocities, (S ) + (S ) , with

(Su )j de a l , the ideal burning velocity 11

2. Peclet constants for "internal wall quenching" of methane-air

mixtures by inert powders 16

3. Peclet constant for "internal wall quenching" of coal dust-air

Generated on 2015-01-31 00:26 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

mixtures by excess coal loading 17

Generated on 2015-01-31 00:26 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

THE THEORY OF FLAMMABILITY LIMITS

Conductive-Convective Wall Losses and Thermal Quenching

by

Martin Hertzberg

ABSTRACT

The concept of limit burning velocities, formulated in an earlier Bureau

of Mines report (RI 8127), is applied to the problem of flame propagation

through tubes of finite size. The limit burning velocity for conductive-

convective wall-loss quenching (process b) is (Su)b = ~-, where the proper

2r0

choice of critical Peclet constant, Pe, is determined mainly by the ratio of

contact-perimeter loss area to flame-zone cross sectional area. This ratio

of loss area to propagation area relates to the shape of the flame front and

its control by buoyancy and by boundary conditions. The comparison of (Su)b

with previously defined limit velocities for systems mixed by natural convec-

tion allows one to assess the influence of tube dimensions and boundary

constraints on the "true" or earthly limits for the three directions of

propagation.

The flame quenching caused by inert powders is shown to be similarly

described in terms of a limit burning velocity, (Su)b . The tube's surface-

to-volume ratio OvL/r0) is simply replaced by the powder's surface area per

unit volume of the flammable mixture. Thermal loss quenching by these

"internal walls" occurs at a critical Peclet constant somewhat higher than

that observed for tubes, and the problem is complicated by the finite heat

capacities of the powders and particle lag effects in the flame front.

INTRODUCTION

In an earlier report (8),2 the concept of limit burning velocities was

used to formulate a quantitative theory of flammability limits. It was shown

that various competing processes can dissipate power from a combustion wave

and thus quench its propagation at some characteristically low limit velocity.

Four competing processes and one complication were involved: (a) free,

Supervisory research chemist, Pittsburgh Research Center, Bureau of Mines,

Pittsburgh, Pa.

Underlined numbers in parentheses represent items in the list of references

Generated on 2015-01-31 00:26 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

at the end of this report.

buoyant convection; (b) conductive-convective wall losses; (c) radiation;

(d) selective, diffusional demixing (the complication); and (e) flow gradient

effects (flame stretch). It was shown that for premixed gases, the "normal,"

upward limit of flame propagation, as conventionally measured, involves

extinction caused by the interaction of processes a and e.

For upward flame propagation, extinction is caused by the ascendance of

the buoyancy force operating through the mechanism of flame stretch. An

upward propagating wave, emanating from a convectively rising flame kernel,

must propagate into cold gas whose motion is also influenced by buoyancy.

Near the top of the flame kernel, the cold surroundings must move outward,

parting from the upward path of burned-gas kernel. In the equatorial regions,

the motion of the surroundings is downward, as required by the net buoyancy

force couple. Upward propagation thus occurs into a velocity gradient in the

cold gas, and at some low, but finite, propagation velocity the flame is

blown off by its own buoyancy-induced flows. For horizontal propagation,

the limit burning velocity is simply derivable as the balance between the

competing combustion and buoyancy forces, and is (Su )a = [2ag pb/pu]1'3.

For upward propagation, it was shown that the blowoff limit velocity was

(Su)a ei ~ ^ (Su)a . These formulae were shown to properly predict both the

compositional dependence of the lean limit for a variety of fuels and the

measured variations of the limit with increasing gravitational acceleration.

Thus it was shown that the very existence of "normal" limits of flammability

at finite fuel concentrations is caused by the competition between combustion

and buoyancy forces, and that this competition results in the presence of a

discontinuity in the burning velocity at the limit. Flame propagation can

occur only if the "true" or ideal burning velocity exceeds the limit velocity.

Compositions whose ideal burning velocities are less than the limit velocity

are quenched by buoyancy and thus have zero values for their real burning

velocity.

The existence of this discontinuity, or minimum (limit) value for the

burning velocity, was shown in a subsequent publication to have a profound

effect on the actual structure of diffusion flames (11) . The concept led to

a new view of diffusion flame structure which is at variance with the tradi-

tional one but which is much more consistent with recent data (15).

This report moves on to consider the second competitive process:

(b) conductive-convective wall losses. An equation is derived for the limit

Generated on 2015-01-31 00:26 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

burning velocity for conductive-convective wall loss quenching, (Su)b.

This equation is equivalent to the concept of a critical Peclet number for

flame propagation through tubes, which concept has already been shown to be

consistent with measured values for quenching diameters. The magnitude of

(Su)b relative to (Su)a allows one to assess realistically the influence of

apparatus dimensions on "true" limit measurements. With further development,

the formula is made applicable to the problem of the thermal quenching of

flames by inert, powdered particles. Such particles can be considered as

"internal walls" contained within the flammable mixture.

THE CONDUCTTVE-CONVECTTVE WALL-LOSS LIMIT: QUENCHING DIAMETERS

Flames propagating through tubes of finite dimensions will lose combus-

tion energy to their surroundings by heat conduction through the tube walls

(since the walls are initially colder than the burned gas). This loss process

generates steep temperature gradients in the gas near the wall and a quenched

boundary layer, as depicted in figure 1. For large-diameter tubes, the

quenched boundary layer is far removed from the central regions and does not

significantly influence the actual burning velocity in the central regions.

However, as tube diameters diminish, the radial temperature gradients in the

quenched boundary layer converge inward and soon begin to influence the bulk

propagation rate. Eventually propagation is quenched at some finite tube

size, referred to as the quenching diameter. These nonadiabatic loss proc-

esses involve the heat flow vectors perpendicular to the propagation direction.

As indicated in figure 1, the losses not only are in the "rim" regions where

the flame zone contacts the wall but also come from the burned gas regions

just behind the flame zone.

The problem has been studied by von Elbe and Lewis (21), Spalding (20),

Friedman (5), Potter and Berlad (17), and Gerstein and Stine (6), and the

results of these and other studies have been summarized in a survey by

Potter (16). The governing equation may be obtained by the following elemen-

tary derivation. For an ideal, flat, laminar, flame-front that is propagating

in steady-state, the combustion power density in the propagation direction is

Loss power density

Quenched

boundary layer

Source power density

Suc/>(Tb-Tu)

-Additional perimeter

area, (/3-D Ac

Rim contact perimeter

area,Ac=2TTr0.AX

FIGURE 1.

- Source and sink power densities for the wall-loss quenching of a flame

Generated on 2015-01-31 00:26 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

propagating in a tube.

(Su)ideal cp [Tb-Tu]. This combustion power density3 is, in effect, the

energy feedback rate from the burned to the unburned gas that maintains the

steady-state propagation rate in a coordinate frame moving at the burning

velocity. The combustion power density is approximately equal to the rate

of generation of chemical enthalpy (by the combustion reactions) per unit

area of flame zone. As shown in figure 1, this power density is axial and

maintains steady-state propagation along the tube axis.

In the presence of cold walls, process b competes with the propagation

process. There is a radial conductive-convective heat flow density to the

tube wall that is given by (Tb -Tu ) = (Tb-Tu) Nu. This heat flow is the

r0

rate of loss of sensible enthalpy from the burned gas per unit area of tube

wall. Now for the axial propagation of a planar flame front in a circular

tube, the area of unburned gas that is activated by the combustion power

density is simply the cross-sectional area, irr0 . For the competing radial

heat loss to the tube wall, the area involved is the flame-zone contact

perimeter area 2irr0Ax plus an additional perimeter area in the burned gases

just behind the flame zone. The total area involved for the heat loss flux

density is taken as 2ir3r0Ax, where g is a dimensionless, geometric wall loss

factor.

The total axial combustion source power is therefore

(S ). , , cp (T -T ) irr 2 .

u ideal b u 0

The total competing radial heat loss power is

(Tb~Tu) Nu 2iT3r0Ax-

The limit velocity for process b is obtained when the ideal combustion

source power is dissipated as heat loss power. Equating source power and

sink power and setting Ax = a/(Su). deal and ot = , gives

(Su)ideal . = (Su)b = (2gNu)1/2 a/r0. (1)

Since the measurable quenching diameter, dq , at the limit of flame propa-

gation through tubes of decreasing diameter is simply 2r0 , and the limit

occurs when (Su)ideal = (Su )b , one has:

dg =

The quantity S d /a is recognizable as a critical Peclet number, Pe, and

this is precisely the dimensionless quantity that correlates the measured

Generated on 2015-01-31 00:26 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

3General introductory concepts and a definition of (Su)ideal are presented in

RI 8127 (8).

quenching diameter data with the combustion and transport properties of the

flammable gas (16-17) . However, it should be recognized that one is not here

dealing with a Peclet number in the conventional sense. The conventional

Peclet number is the ratio of convective to conductive heat transfer to a

given surface. Here, instead of a single surface, there are different

"surfaces" that are approximately orthogonally oriented relative to one

another. The "convective flow" reflected in Su [or (Su)ideal ] relates to

energy feedback rate from burned to unburned gas that is necessary to main-

tain the steady-state propagation rate at Su. It is the flow "eigenvalue"

for steady-state flame propagation. The conductive wall loss reflected in

a/dq is orthogonal to the direction of flame propagation. As indicated in

the derivation, in the case of equation 2, the Peclet constant is the ratio

of energy transfer across the flame-zone source (required to maintain propa-

gation at Su) relative to the rate of energy transfer in the orthogonal direc-

tion through the quenched boundary layer at the wall. The mechanism of energy

transfer is pure conduction and diffusion for both directions. A velocity Su

is involved, but there is really no convective heat being transported to any

solid, physical surface by that flow.

In this idealized derivation of equation 2, the axial combustion source

power for laminar, adiabatic flame propagation was compared with the radial

heat loss power to the wall. In reality, as one approaches the limit, flame

curvature appears, propagation becomes nonadiabatic, and axial heat losses

appear from the flame front to the cooler unburned gases. Thus in a system

that is actually undergoing quenching, the orthogonality between propagation

and loss directions begins to dissappear. Some of these geometric complexi-

ties will be touched on shortly. Nevertheless, their effect is quantifiable

by a proper choice of the 3-factor.

With these uncertainties in mind, equation 2 indicates that steady-state

propagation is possible only if the Peclet number exceeds some critical value;

that is, if the axial combustion source power exceeds the radial power loss.

Spalding's solution (20) for the critical Peclet number is 60 for circu-

lar tubes, and this agrees with the earlier calculations of von Elbe and

Lewis (21). Friedman (5) measured Pe values of 14 to 18 for downward propaga-

tion through plates. For plates, the ratio of loss area to flame area is

half that for tubes, and hence the critical Peclet constant should be 0.71

times as large. Thus Friedman's plate values would correspond to tubular Pe

Generated on 2015-01-31 00:27 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

values of 20 to 25. It should be recognized that measured values are gener-

ally obtained for relatively short channels of finite length, whereas the

theories are steady-state calculations that apply to tubes of infinite length.

Since shorter tubes are less effective quenchers than tubes of infinite

length, it is not surprising to find the measured quenching diameters sys-

tematically smaller than the theoretical ones. Other factors involving the

flame front curvature and flow convergence at the tube inlet should also

influence the measured value of the critical Peclet number for wall-quenching.

Friedman verified the essential validity of equation 2 by varying Su and a

for hydrogen-oxygen flames containing various inert gases. Potter and

Berlad (17) verified it for a wider range of fuel-oxidizer combinations and

also showed that the pressure dependence of d was accurately predicted by

the pressure dependence of the ratio a/Su, as equation 2 requires.

As indicated in the derivation, the absolute value of the critical Peclet

number is controlled by the ratio of the contact perimeter loss area to the

flame-zone cross sectional area. For the assumed planar flame front in a

tube of circular cross section, that ratio is 2gAx/rQ = 23a/r0Su. Note that

this ratio of loss area to source area varies inversely with Su. An exact

choice of the perimeter area factor, g, is difficult to make. Its accurate

determination requires a detailed solution to the flame propagation equation

in three dimensions. This uncertain factor is thus lumped into the Peclet

constant, which is normally determined empirically.

One must recognize, however, that the above ratio is necessarily influ-

enced by variations in flame zone shape, and that the overall balance may be

influenced by flow convergence and convection effects. Flame shape is con-

trolled by buoyancy forces, radial boundary constraints (tube shape), axial

boundary constraints (closed versus open-ended ignition), and selective dif-

fusional demixing effects (cellular flame structures). These processes must

exert some influence on the critical Peclet number at the wall quenching

limit. These factors influence the precise details of the flow structure

on both sides of the flame front and near the walls, and they may add a flame-

stretch factor to the problem.

BUOYANCY AND FLAME SHAPE FOR NEAR LIMIT FLAME PROPAGATION

Observations of the propagation behavior of near-limit mixtures clearly

reveal significant differences in both the flame and the flow structure for

the three directions of propagation. Levy's studies (13), for example, showed

that the flame-front curvature is spherical during upward propagation in tubes

and that the curvature tends to be maintained even during flame quenching; but

during downward propagation in tubes, the curvature diminishes, and at extinc-

tion the front is essentially flat. The recent observations of Sapko, Furno,

and Kuchta (18) of the structure of near limit flames in a very large scale

apparatus clearly shows that the same buoyancy-induced distortions occur in

the absence of wall-boundary constraints. In the composition region between

the upward and downward limits, the "normal" spherical flame is replaced by

a severely distorted fireball shape.

If one considers an initially spherical flame kernel and assigns to it a

uniform, upward buoyant acceleration vector (8-ll_) , the observed distortion

is precisely what one would expect. That is, the buoyant acceleration vector

is normal to the flame front at the top of the sphere, and hence the entire

Generated on 2015-01-31 00:27 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

vector magnitude accelerates the upward propagation velocity (neglecting the

flame-stretch factor for the moment). At other points in the upper hemisphere,

the buoyant acceleration is diminished by the cosine of the angle between the

surface normal and the upward-directed, buoyant acceleration vector. A simple

vector construction of the resultant flame-front displacement shows that the

net effect of buoyancy is to maintain spherical curvature in the upper hemi-

sphere as the spherical flame kernel expands and rises.

For downward propagation, on the other hand, the buoyant acceleration

vector is directed opposite to the outward normal propagation vector at the

flame front. Buoyancy thus decelerates propagation in the lower hemisphere

of the flame kernel. At the very bottom of the sphere, the entire vector

magnitude of the buoyant deceleration reduces the downward propagation veloc-

ity. At other points in the lower hemisphere the deceleration is diminished

by the cosine function. The net effect for downward propagation is to reduce

curvature and to generate a flattening flame-zone, for as the flame kernel

expands, its rate of descent tends to be canceled by its buoyant deceleration.

This elementary viewpoint generates approximately the correct overall flame

shape for free-space propagation, even to the extent of reversing the curva-

ture and causing the flame front to dimple inward at the bottom of the sphere

when the buoyant velocity begins to exceed the gravity-free flame speed (18).

For tubular propagation, such changes in flame shape must necessarily

influence the critical Peclet constant. Consider first propagation from the

open end of a tube after planar ignition. Neglect buoyancy for the moment

and allow the wave to propagate at Su into the initially static, cold gas.

Consider the details of the flow structure generated by the combustion force

field at the flame front, as it interacts with the wall boundary constraints.

This flow structure is described by Jost (12), who showed that propagation of

the flame front as a flat, planar wave soon becomes impossible. The gas

entering the flame front from the unburned side must adjust to a Pouiseille

velocity profile on the burned side, and it must also uniformly accelerate to

the velocity, Su (pu/pb), in the direction perpendicular to the flame front.

It becomes physically impossible to satisfy both constraints across a flat

flame surface, and accordingly the flame front takes on a paraboloid shape in

steady-state propagation, and vortices appear in the cold gas zone just ahead

of the flame. This generation of flame curvature is independent of buoyancy.

Some of the additional complexities that are observed in the presence of

buoyancy are illustrated in figures 2 and 3, which depicts propagation in a

horizontal tube. The ini-

tial planar wave must not

only adjust to the flow

Su Unburned gas

Burned gas

Resultant flow

Combustion force

2a

Normal

jCold

Generated on 2015-01-31 00:27 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Near-limit

Bouyancy force couple

FIGURE 2. - Flow trajectories in combustion and buoyancy

force fields.

constraints just indicated,

but it must also respond to

the buoyancy force couple.

In the absence of buoyancy,

cold gases enter the flame

front at Su (fig. 2, top),

and the gases are continu-

ously heated and acceler-

ated to Sb as they react.

The resultant expansion

generates a flow through

the open end of the tube.

For normal rapid flame

propagation, the combustion

force dominates, and the

neglect of buoyancy is

justified.

case

For the near-limit

, however, buoyancy

Planar

ignition

Maximum

velocity

<\ \ y

X\\ Convective

<xXx>'vortex

N/'

FIGURE 3. - Flame shape and flame-zone structure for

horizontal propagation in a tube.

cannot be neglected. In the

near-limit case (fig. 2,

bottom), the trajectory of

a small volume of gas is

depicted as it approaches

the flame front and reacts

under the simultaneous influ-

ence of both combustion and

buoyancy forces. The flow

of reacting gases is toward

the open end and upward

toward the top of the tube.

Since buoyancy is a force

couple, this upward flow is

accompanied by downward flow of cold gas (9-11). This convective cell motion

is superimposed on the developing paraboloid flame shape, as depicted in fig-

ure 3. The convective vortex tends to accumulate hot, burned gases near the

top of the tube and cold, unburned gases near the bottom. The buoyancy vortex

thus maintains (and enhances) propagation near the top and diminishes the rate

of propagation near the bottom.

The resultant changes in the shape and width of the flame front are

depicted in figure 3. As the horizontal limit is approached, a condition is

reached in the bottom regions where the component of the buoyant retardation

velocity perpendicular to the flame front just balances the true flame speed.

Propagation will then cease in the lower region, and the flame is said to

"attach" to the top of the tube. At the horizontal limit, flame propagation

Generated on 2015-01-31 00:28 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

of the "attached" flame near the top of the tube is finally quenched by

conductive-convective wall losses and by flame-stretch flow gradient effects.

The radial variation in flame-zone thickness that might be expected near

the horizontal limit is also depicted in figure 3. The flame-zone thickness

in the lower regions is widened by convective and diffusive flows near the

stagnant flame front. There is a flame zone widening near the top of the tube

because of losses to the quenched boundary layer at the wall. The maximum

horizontal propagation rate should be just below the top of the tube, where

the flame zone is depicted as narrowest. There is necessarily a flow gradient

that is induced by the convective vortex in all regions of the flame zone.

The limit velocity is obtained by balancing the buoyancy force couple that

generates the convective vortex against the combustion force (8).

THE INFLUENCE OF APPARATUS DIMENSIONS AND BOUNDARY CONSTRAINTS

ON THE MEASURED LIMITS OF FLAMMABILITY

Some of the lean limit measurements for methane-air mixtures in tubes of

varying size are graphically summarized in figure 4. Also shown are the burn-

ing velocity measurements of Andrews and Bradley (1), Gunter and Janisch (7),

Edmondson and Heap (3_), and Egerton and Thabet (4_) . The lean composition

limits for upward propagation are shown as upward-directed arrows. The arrow

lengths are proportional to the diameters of the tubes within which the limits

were measured.

u = Upward propagation limits

hr Horizontal limits

d = Downward limits

d 7.5 d5.0d2.5

IiiIIiiiII

5.0

6.0 6.517.0 8.0 9.0

METHANE, pet

10.

FIGURE 4. - Lean limit measurements in methane-air

mixtures for the three directions of flame

propagation in tubes of varying diameter,

and the measured burning velocities at

those limit concentrations.

The limits for downward

propagation are shown as

downward-directed arrows,

and the horizontal limits

are shown as short, vertical

lines with horizontal arrows

running through them. The

downward-directed arrows

marked d 7.5, d 5.0, and

d 2.5 are the composition

limits for downward flame

propagation in tubes of diam-

eters 7.5, 5.0, and 2.5 cm,

respectively. The horizontal

limits are marked h, and the

upward limits are marked u.

These data were taken from

the compilation on Linnett

and Simpson (14_). The limits

marked u' are the more

recent measurements of

Generated on 2015-01-31 00:28 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Levy (13). The limits marked

u'r are the data in 25-cm-

diam tubes, compiled by

Coward and Jones (2_) . The

leaner value is for propaga-

tion from the closed end;

the richer value is for

propagation from an open end.

The difference may be attrib-

utable to turbulence, to

subtle differences in the

criteria chosen for the limit

in the two cases, or to some

other unexplained effect.

Other data are summarized

by Coward and Jones.

The "accepted" value is 5.0 percent (^2). This is presumably the lean

limit for upward propagation, in the absence of wall complications, using a

relatively "soft" propagation criterion. The soft criterion is the detection

of some kind of contiguous flame propagation that extends far beyond the

region of ignition.

From equation 1, one may estimate the limit burning velocity for wall

quenching. For a 2.5-cm-diam tube, using a consistent value (8) for a of

0.55 cm2/sec and setting Pe = 25 (its measured value) gives (Su)b = 6 cm/sec.

This value is about equal to (Su)a , the horizontal limit velocity for quench-

ing by natural convection. One would thus conclude that wall losses in a

2.5-cm-diam tube are of comparible significance to buoyancy in determining

the horizontal limit. But one may go a bit further. Consider the measured Su

10

data of Gunter and Janisch (the upper curve in figure 4) and the measured

horizontal limit of 6.0 percent methane in the 2.5-cm-diam tube (h 2.5 in

figure 4). Projecting that limit composition upward until it intersects the

S curve gives a limit burning velocity of 12 cm/sec. Is it a coincidence

that this value is just equal to the sum of (Su)a and (Su)b? Note also that

the average difference in Su for the u and h values in tubes of fixed diameter

is close to the difference of 3 cm/sec between (Su)a>e.f and (Su )a . One is

thus tempted to postulate a simple additive effect: in the presence of both

buoyancy and wall losses, the horizontal limits would correspond to the

condition

(Su)ideal = (S). + (Su)b (3)

For the upward limit one would replace (Su )a by (Su)a,e^.

The extent to which equation 3 agrees with the data is indicated in

table 1. The measured lean limits in tubes of varying diameters are taken

from Coward and Jones (2). (Su)b is then calculated for the varying tube

diameters by setting Pe = 25, as above. The measured limit composition gives

a value for (Su)ideal , which is taken from the burning velocity data of

Gunter and Janisch (7_) or Egerton and Thabet (4). The (Su )a value for

horizontal propagation is taken as 6 cm/sec, while the (Su)a e/|. value for

upward propagation is taken as 3 cm/sec. Comparison of the last two columns

in table 1 shows that equation 3 is approximately correct. However, one can-

not be too impressed with the precise values for each case considered in view

of the expected experimental uncertainties. For example, the measured limit

compositions include a range of water vapor contents, varying from methane-

air mixtures that were dry to mixtures that were fully saturated at ambient

conditions. Some of the limit compositions were for closed-end ignition,

others for open-end ignition. The measured Su values are taken to be equal

to the ideal ones. This is a good assumption for mixtures above 7 percent

methane, but true burning velocities are exceedingly difficult to measure in

the range of 5 to 6 percent methane, precisely because of these competing

processes.

In any case, it seems clear that in order to obtain accurate lean limit

measurements that are independent of wall loss effects, one must satisfy the

condition (Su )b (Su )a or (Su)a>e^. If, for example, one sets (Su )b = 0.1

(Su)a to satisfy the above requirement, then the flammability tube for hori-

Generated on 2015-01-31 00:28 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

zontal limit measurements should have a tube diameter in excess of 20 cm.

However, the absolute error introduced by using a 10-cm tube is at most 0.1

to 0.2 percent methane: a relative error of 1 to 2 parts in 50. Clearly,

limits measured in larger diameter tubes are "truer" earthly limits than those

measured in small-diameter tubes. Ultimately, the size required to obtain the

"true limit" is determined by the accuracy desired by the investigator.

11

TABLE 1. - Comparison of the sum of limit velocities, (Su)a + (Su)b,

with (Su)j deal , the ideal burning velocity

Tube

diameter,

cm

Measured lean

limit (2) ,

pct methane

(S>b.

cm/sec

cm/sec

cm/ sec

HORIZONTAL LEAN LIMIT

7.5

5.4

5.0

5.7

2.5

5.8

11

12

.90

7.8

15

30

21

.56

8.4

35

31

.45

110.0

31

Generated on 2015-01-31 00:28 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

25

4 2 (max)

37

.36

(2)

38

42 (max)

44

UPWARD LEAN LIMIT3

10.2

5.0

7.5

5.4

4.0

5.8

2.5

5.5

10

12

would certainly be turbulent. This could alter the proportionality constant

in equation 4 quite significantly. Nevertheless, for smaller dimension, near

limit velocities, and in the absence of upper boundary constraints, equation 4

should control the downward stagnation limit. As before, the limit is given

by the condition (Su)ideal = (Su )a ^ + (Su )b . Now, however, the limit veloci-

ties for both process a and process b contain an r0-dependence. (Su)b varies

as l/ro as before, but now (Su )a depends on r0 . The sum (Su)a^ + (Su )b now

has a minimum in its r0 dependence. This is shown in figure 5, where the sum

of limit velocities for downward propagation is contrasted with the sums for

upward and horizontal propagation. In the latter two cases, as tube diameters

increase one approaches a "true" limit condition, and the limit velocity

becomes completely independent of tube dimensions and depends only on

process a.

For downward propagation, the situation is clearly more complicated.

There is no asymptotic value for large r0, there is only a broad minimum in

the region d0 ^ 6 to 24 cm. This transition region is in fact the region

where most of the measurements of downward limits are made. For the curves

drawn in figure 5, an

attempt is made to correct

the (Su)b values for the

effect of flame front curva-

ture. The flat flame front

for the downward propagating

wave should be more easily

quenched by wall losses than

the curved flame fronts for

the upward, or horizontally

propagating waves.

Thus for downward prop-

agation a larger critical

Peclet constant was used:

namely, the theoretical

value of Pe = 60. For

horizontal and upward prop-

agation the measured value

Generated on 2015-01-31 00:29 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

of Pe = 25 was used for

evaluating (Su)b. Note that

the calculated minimum in

the sum of limit velocities

for downward propagation is

about 9 cm/sec, and that

this minimum occurs at

d0 = 11 cm. That velocity

is close to the measured Su

value that one would obtain

from figure 4 by estimating

the d 11 limit as 5.8 pct

methane. That 5.8 pct

Horizontal

(Su)Q+(Su)b

(Su)a,ef+(Su)b

i i i I 11

0.5 I 5 10

TUBE DIAMETER,d0,cm

FIGURE 5. - The sum of limit burning velocities for

(a) buoyant-convective quenching, and (b) wall-

loss quenching, as a function of tube diameter for

the three directions of flame propagation.

13

methane value is indeed about the leanest value observed for the downward

propagation limit.

However, the curve in figure 5 predicts that downward limits should shift

to richer values as dimensions increase for d0 > 20 cm. This is in marked con-

trast to the asymptotic behavior for the upward and horizontal curves.

Although there are reliable data to indicate that the formula for the buoyant

rise velocity of an expanding flame kernel, which was used to derive equa-

tion 4, is correct (9), there are virtually no data on the downward flame

propagation limits in tubes of diameters significantly larger than 20 cm.

Equation 4 and the downward curve of figure 5 predict that downward lean

limits would become richer for these larger dimensions. For example, the

sum of limit velocities in a 50-cm-diam tube is 13.5 cm/sec. From figure 4,

one would predict a downward limit of 6.2 percent methane for the 50-cm-diam

tube. For a 100-cm-diam tube, one predicts a downward limit of 6.5 pct meth-

ane. And as indicated earlier, the downward limit should reach the stoi-

chiometric value of 9.7 pct in a tube 1,000 cm in diameter.

A critic would be tempted to claim that such projections are unrealistic.

They appear to be artifacts of buoyancy. They result from the operational

definition of downward flame propagation. The tube is necessarily constrained

to be at rest in the laboratory reference frame as the center of flame prop-

agation rises at the buoyant velocity. Yet, to an observer at a large dis-

tance below the point of ignition, the effect is real in that the flame front

would never reach him if the combustible medium is of infinite volume and if

the top of the container is open to the surroundings.

It is this later condition that suggests that the effect is indeed arti-

ficial; for if the top of the container is closed, then regardless of the

point of ignition, the buoyant rise velocity will necessarily approach zero

as the upward propagating wave approaches the closed top of the container.

Once the burned gas column can no longer rise by buoyancy, downward propaga-

tion becomes possible, and the flame front can then reach an observer situated

below the point of ignition. Such an effect is indeed observed for propaga-

tion in a 12-ft-diam sphere.

Consider the motion and shape of a fireball in an N2-diluted methane-air

mixture, whose measured burning velocity of about 9 cm/sec corresponds to that

of a 5.8 pct methane-air mixture. Its motion and shape are depicted by Sapko

Generated on 2015-01-31 00:34 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

and coworkers (18, fig. 6).1+ The mixture does not propagate at all in the

downward direction once the fireball diameter exceeds 10 to 20 cm. This

causes the fireball shape to distort markedly and to dimple inward at its

bottom. The center of mass of the distorted fireball continues to rise upward

at its buoyant velocity until it approaches the closed top of the spherical

container. As soon as it contacts the top of the sphere, its buoyant velocity

decelerates to zero, and propagation in the downward direction becomes possi-

ble. The fireball is then observed to establish a flat flame front that

4The actual mixture was a ternary one containing 6.9 percent CH^j , 65.8 percent

air, and 27.3 percent N2, which is equivalent to a binary mixture of 5.8

percent CH4 in air.

14

finally does propagate downward and completely consumes the remainder of the

flammable mixture within the 12-ft-diam sphere.

Clearly the curve for downward propagation, (Su)a^, + (Su)b^., in figure 5,

is calculated on the assumption that the flammability tube is open so that

there is no impedence to the cold gas flows required to compensate for the

buoyant rise velocity of the burned gas column. As the burned gas column

rises in an open tube, cold gas exits from the top and enters from the bottom.

Clearly if the tube is closed at the top, equation 4 soon becomes inapplica-

ble, and the downward limit can no longer be a stagnation limit. If the tube

is closed at the top, and the gas is ignited at that closed end, then the cold

gas flow structure ahead of the wave is driven by the burned gas expansion

process, the boundary condition at the top of the tube, and the required

Pouiseille or turbulent velocity profile in the cold gas venting from the

bottom. In that case, the downward limit would presumably be governed by

flow gradient effects and flame stretch. The sum (Su )a ^ + (Su)b would then

probably approach an asymptotic limit in the 6-cm/sec range.

In this sense, the buoyancy effect for downward flame propagation appears

to be an artifact of the experimental method: the measured downward limit

becomes critically dependent on whether the system is open or closed with

respect to the surroundings. But it seems hardly correct to consider the

effect of buoyancy as an artifact, when it is precisely the process responsi-

ble for the very existence of the limits of flammability in the first place.

THERMAL QUENCHING BY INERT POWDERS

Return for a moment to the problem of flame propagation in narrow tubes,

and the derivation of equation 1, which defines the limit burning velocity

for quenching by heat losses to the tube walls. It is given by:

where Pe is the critical Peclet number for wall quenching. Now for a circular

tube of length and radius r , the internal tube surface area to which com-

bustion energy is lost is 2irr . The source of combustion energy is the gas

contained in the tube, whose volume is irr02. The surface to volume ratio is

therefore A/V = 2/r0 , and it is this ratio that determines the critical dimen-

sion at which the exothermic source strength is overcome by the thermal losses

to the boundary. In terms of this surface to volume ratio, the limit burning

velocity becomes:

(A/V)

Generated on 2015-01-31 00:34 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Now suppose that the wall boundary is not an external one but an internal

one. Imagine that the wall is pulverized into a powder in such a way that the

tube's surface-to-volume ratio is preserved in the surface area of powder per

unit volume of flammable gas. In this sense, the quenching effect caused by

15

the addition of an inert powder to a flammable mixture may be viewed as the

thermal quenching of a combustion wave by "internal walls."

One should, however, be cautious in applying this concept, for there is

a significant difference between the internal surface area of the powder and

the surface area of a contact wall. The wall boundary can conduct heat to a

reservoir of essentially infinite heat capacity; however the "internal wall"

or powder has a finite heat capacity. We will evaluate the finite heat capac-

ity effect and related questions in the following section.

With this precaution in mind, it is assumed that these inert powder walls

can quench the flame by conductive-convective wall losses in the same way that

external walls can. Consider an inert dust of solid density p , consisting of

spherical particles of diameter dp , dispersed in the gas mixture at a mass

concentration of dust, Cm. The number density of particles is n = 6Cm/irdp3 pp .

The surface area of dust per unit volume of gas is mrdp2 , and hence the sur-

face area of powder per unit volume of flammable gas is

.,.. , 2 6xlQ-2 Cm (mg/1)

A/V = rnrd z = -- - -. r (7)

!> Pp dp (ym)

Substituting equation 7 into equation 6 and setting a = 0.55 cm /sec gives

f* M 9< in

CS ) . = 8.25 x 10

,.

u b Pp dp (ym)

The quantity of (Su ) 'b is now the limit burning velocity for quenching

by pure thermal losses to inert particles in the flammable mixture.

With the several cautions in mind (to be discussed in the following sec-

tion) , one can use (Su ) 'b as the limit burning velocity for inert particle

quenching. In addition, because of the additivity principle (Eq. 3), (Su)'b

may also be viewed as the reduction in the ideal burning velocity caused by

the presence of the inert powder in the flammable mixture. A system is com-

pletely quenched by the inert powder when that reduction in burning velocity,

plus the reduction caused by buoyant quenching, is equal to the ideal value;

that is, when (Su ) 'b + (Su )a = (Su)ideal . However, if the flame is not

quenched, the presence of thermal losses to the inert additive is still felt

by the propagating flame, and its real burning velocity is reduced by (Su ) 'b .

(9)

Generated on 2015-01-31 00:34 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Differentiating equation 8 with respect to Cm gives

d(Su)'b 8.25 x 10~3 Pe

dCpd

mpp

Equating the increase in (Su) 'b for a given Cm with the reduction in

measured burning velocity S , and solving for Pe gives

Pe . -

ACm 8.25 x 10~3

16

Smoot and Horton (19) have studied the effect of added rock dust (CaC03)

and Alumina (A1203) powders on the burning velocity of methane air mixtures.

Their data are used in table 2, and the resultant values for Pe calculated

from their data and equation 10. The addition of the powdered inert dusts

results not only in thermal quenching by heat conduction to the additive, but

also a considerable increase in emissivity of the burned gas caused by the

addition of the dust. This increase results in radiative losses, which are

describable in terms of a radiative loss limit velocity (Su)c . The magnitude

of (Su)c is between 2 and 8 cm/sec depending on the dust concentration and

flame temperature. Since both loss effects are present simultaneously, it

is necessary to subtract this small value of (Su)c from the measured ASU

values in order to isolate the effect that is attributable to (Su)b alone.

The resultant Peclet numbers for inert quenching average to about Pe = 100.

Although this value is only somewhat higher than the theoretical wall loss

quenching value of P = 60, one cannot be too impressed with the result that

the 20ym alumina particle was no less effective than the 9um particle.

TABLE 2. - Peclet constants for "internal wall quenching"

of methane-air mixtures by inert powders

Particle diameter,

Methane

in air,

pct methane

Measured ASu/ACm,

cm/sec

ds , pm

Pe1

mg/1

CaCO, (pd = 2.71)

10

10

10

9.7

8.0

12.0

0.024

.023

80

96

77

Generated on 2015-01-31 00:34 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

.029

A1203 (pd = 3.97)

9.7

8.0

9.7

0.018

.021

.018

78

92

173

20

Calculated from equation 10.

Another example of the effect of internal quenching by inert particles

may be obtained from the Smoot and Horton (19) data on the burning velocity

of coal dust-air flames. Their data show that burning velocities peak at

coal dust concentrations that depend markedly on particle size. For fine

dusts, the maximum burning velocity, (Su) (max), occurs at relatively low

dust concentrations, whereas for coarser dusts the maximum appears at a much

higher dust concentration. (Su) (max) is significantly smaller for the

coarser dusts, which require higher dust loadings.

The reason for this behavior relates to the fact that the overall rate

of flame propagation at these maxima in S , is limited by the rate of pyrol-

ysis or devolatilization of the coal particles. All peak values correspond

to the consumption of only a stoichiometric concentration of combustible

volatiles (10); however, the coarser the dust, the more difficult it is for

the entire particle to devolatilize during its passage through the flame

front. The coarser the particle, the smaller is the fraction of its com-

17

mass concentration of dust required to generate a stoichiometric concentration

of combustible volatiles. For the coarse coals, only the shallow shell near

the surface or at the sharpest corners can devolatilize in time to contribute

to flame propagation. For the finer coals, most of the coal particle devolati-

lizes in time to contribute.

In either case, the excess mass of coal (or the remaining char residue)

may be viewed as an inert powder to which combustion energy is lost. The

effect of this excess dust loading is seen in the data as a reduction of

Su(max). Table 3 presents the Peclet constants, calculated from equation 10,

associated with the reduction in Su (max) by the excess inert dust. For the

coal one uses

pp =1.4 g/cm3.

TABLE 3. - Peclet constant for "internal wall quenching" of

coal dust/air mixtures by excess coal loading

Particle

diameter,

ds , ym

Su (max)

cm/ sec

Cm (max)

mg/1

Measured ASu/ACm,

cm/ sec

Pe1

mg/1

7.3

12

30

32

250

400

1,500

0.0588

.0406

.0186

83

95

27

10

Generated on 2015-01-31 00:35 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

73

Calculated from equation 10.

The average constant for the quenching of excess coal dust in the

coal/air flames is Pe = 84, which is quite close to the values calculated

in table 2 for the quenching of methane-air mixtures by rock dust and alumina

powders.

FINITE HEAT CAPACITY EFFECTS; PARTICLE LAG IN THE FLAME FRONT;

THE PARTICLE BOUNDARY LAYER

A caution was indicated in the derivation of equation 8, that related to

the finite heat capacity of an added powder. Equation 6, for tube-wall

quenching, was derived under the assumption that heat was continuously being

conducted through the boundary layer and through the interior surface to a

medium of infinite heat capacity. In effect, it was assumed that the cold

wall, initially at Tu, remained cold at Tu during and after the passage of

the combustion wave. This assumption was included in the subsequent deriva-

tion of equation 8. If, however, the wall boundary has a finite heat capacity

(relative to the burned gas volume it contains), then the temperature gradient

TT

-; and its associated heat loss to the wall would not be maintained.

This is the case for the "internal walls" of inert particles, where their

finite heat capacity may cause their wall temperature to rise significantly

above T during the passage of the flame front. If the particle mass is so

small that its temperature rises as fast as the gas temperature rises within

the flame front, then thermal losses are not maintained, and the gradient

T-T

bu

approaches zero. In that instance it is not the surface-to-volume

18

ratio of the dust that is important, but only its heat capacity per unit vol-

ume, which is proportional to the mass concentration Cm, and independent of

the particle diameter.

The extreme case of such a "small particle" is the case of an inert gas

such as N2. It is instructive to consider this extreme case in order to be

aware of the limitations of equation 8. The N2 particles added to a flammable

mixture are heated to Tb as rapidly as the combustion products, and since they

are not maintained at Tu, using equation 8 to calculate inerting requirements

would produce an absurd result. This can be demonstrated as follows:

One could argue that since the inert gas is mixed on the molecular scale,

its particle diameter d should be the molecular diameter of the nitrogen mol-

ecule. Using the liquid (or solid) N density for p and typical values for

Pe and (Su)ldeal would lead to the prediction that aPstoichiometric methane-

air mixture would be inerted at a Cm value of only 1 mg/1. The prediction is

clearly erroneous, since the N2 molecules are not cold walls that remain cool

during flame front passage, but rather their heat capacity is so small that

they precisely follow the increasing gas temperature. Data show that the mix-

ture is actually inerted at a Cm of 500 mg/1, which corresponds to some

36 volume pct added N2. The absurd result of 1 mg/1 was obtained because

equation 8 was used beyond its range of validity. The 1 mg/1 value could be

valid only if each N2 molecule had an infinite heat capacity and could thus

remain at Tu even though it was imbedded in reacting gas whose temperature

was increasing to Tb. This is physically impossible, since each molecule has

a very small heat capacity.

The heat capacity per unit volume of 1 mg/1 of N2 is trivial compared

with the heat capacity per unit volume of the reacting mixture whose total

mass concentration is 1,300 mg/1. Thus, the added 1 mg/1 of N2 has a trivial

effect in reducing the flame temperature or the rate of flame propagation.

In reality, it requires some 500 mg/1 of added N2 to reduce the flame tempera-

ture from its adiabatic value of 2,200 K to its limit value near 1,500 K (8).

One obtains some additional insight into the fallacy just considered by

using the measured value of Cm - 500 mg/1 and substituting the typical values

for Pe and Su to calculate the "effective particle diameter" at which quench-

ing occurs if the added nitrogen were to be considered as an inert powder.

The value is d - 3 ym. This value is clearly much larger than the molecular

Generated on 2015-01-31 00:35 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

diameter. But it is also significantly larger than the mean free path for

molecular collisions, L - 0.1 ym. This latter condition is necessary for the

bulk property definitions of density, heat capacity, and thermal conductivity

to have their normal meanings in the fluid continuum. Clearly continuum equa-

tions were used to derive equations 1 and 8, and hence their validity is

limited to the condition d L. This condition is violated by choosing

the molecular diameter of the N2 molecule as the particle diameter.

It is constructive to consider this particle heat capacity and heating

rate problem a bit further. To what extent can inert particles added to a

combustible mixture follow the increasing gas temperatures in the flame front,

or to what extent do they lag the gas temperatures?

19

For an initially cold, inert particle of finite mass subjected to a lami-

nar heat flux I, the rate of temperature rise is given by:

V p c -T = A I

p p p dt P

where V is the particle volume, c its heat capacity per gram. and A is its

cross sectional area. The laminar heat flux to the particle is equated to the

laminar combustion power density, S c p (T - T ). This gives:

- irr 3 p c = iir 2 S cp(T-T).

3 P p p dt P""

The specific heats per gram of particle and gas are approximately equal

so that cp - c. Solving for the characteristic time constant for particle

self-heating in the flame front gives:

Tp = I r" IT k'

For the flame gases, the characteristic heating time is T = a/Su2 .

Hence the ratio of particle heating time to flame gas heating time is

_ ,4 pp_ S^

Tp'Tgas "3" rp ~ ~a~'

If this T-ratio is unity, then clearly the particle temperature should

closely follow the gas temperature. A T-ratio of less than unity is a physi-

cally impossible artifact (it would violate our continuum constraint that

d L). In any case, it too would mean that the particle temperature should

closely follow the increasing gas temperature of the combusting system. In

either case (T-ratio equal to or less than unity), the particle cannot be

considered to be a cold wall, and only its mass or total heat capacity is

significant in its quenching effect. If, on the other hand, the T-ratio is

much larger than unity, then equation 8 is applicable, and the particle should

behave as a cold wall during flame front passage. Setting a = 0.55 cm2/sec,

p /p - 3 x 103 for typical particle and gas densities, and using the limit

condition Su = (Su)a^ - 3 cm/sec, one obtains r - 0.5 ym for a T-ratio of

unity. Thus a lym-diam (or smaller) particle should closely follow the

increasing gas temperature within the flame front, and equation 8 would not

be applicable. On the other hand, for a lOym-diam particle the T-ratio is 10

and hence the rate of temperature rise of the particle will be an order of

magnitude slower than that of the gas. Thus a lOym-diam (or larger) particle

should behave as a cold wall during flame front passage, and equation 8 should

Generated on 2015-01-31 00:35 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

be applicable. The particles considered in tables 2 and 3 were in the appli-

cable size range.

A final complexity involves the kinematics of inert particle motion in

the flame front. It was implicitly assumed in deriving equations 8 and 11

that the particles entering the flame front would instantaneously follow the

accelerating gas flow so that their velocities were always equal to the gas

20

velocity. We now consider the validity of that assumption. The question is

important for two reasons. First, there is the question of the heat transfer

coefficient from the gas to the particle. The conductive-convective heat

transfer rate from a gas to a solid surface (or the reverse) is proportional

to the Nusselt number, Nu, which is proportional to the ratio of particle

dimensions to boundary layer thickness. For the dimensions and velocities of

interest, Nu is proportional to the square root of the Reynolds number, /Re.

For small particles and low relative velocities, Re is small and Nu -> 1. This

condition was implicitly assumed in the derivation of equation 11, and only

pure conduction across a boundary layer whose dimensions were comparable to

the particle diameter, was considered to be important in the heat transfer

process. If, however, the particle should lag the gas flow, Nu can exceed

unity. and there is a larger heat transfer to the particle. If there is

significant relative motion of gas with respect to the particle, the boundary

layer thickness becomes smaller than the particle dimensions. Actually, for

the dimensions and velocities of interest here, Nu is never significantly

larger than unity, and this first problem is of minor importance.

What is significant, however, is the second reason for considering parti-

cle lag effects. If the particle lags the gas flow so that its time of travel

through the flame front is significantly longer than that of the gas, then its

effective concentration in the flame front will be higher than its initial

concentration, Cm, in the static mixture. If particles lag in the accelerating

gas flow, they pile up in the flame front and their effect is magnified.

Consider now the kinematics of particle motion in order to estimate this

particle lag effect. The force on a sphere of radius r , which is imbedded in

a fluid of viscosity n moving with a uniform relative velocity u with respect

to the sphere, is given by Stokes' law:

F = Birr n u

pP

The drag force is in the direction of the relative velocity u. A more accu-

rate formula uses the factor (1 + 3/16 Re), which would increase the drag

force by a few percent in the most extreme case. We here ignore the second

term and use the Stokes' formula. Now in a coordinate system moving with the

flame front at the burning velocity, gas and particles are initially at rest

with respect to each other, and both enter the flame front at the velocity S .

Generated on 2015-01-31 00:35 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

The gas accelerates as it traverses through the flame front and reacts to give

products whose exit velocity is Sb = Supu/pb. The details of the acceleration

under the influence of the combustion force depend on the details of the flame

structure. It is convenient, and probably reasonably accurate, to assume a

simple linear increase in gas velocity as it traverses across the flame front.

Thus, initially at t = 0 (and x = 0), (vg)0 = Su while at t = Tgas = a/Su2

(and x = a/Su), (vg ) = Sb = Supu/pb. This gives

Su - Su

a/S.

t = s,, +

] -1

21

Now the particle's initial velocity upon entering the flame front is also

(vp)0 = Su. Hence initially (u)0 = (vg)0 - (vp)0 =0. At any time t, while

the particle is within the flame front (that is, t < T gas), the relative

velocity is

u=v-v=

sP

S3

_u

pb

t-V

The equation of motion for the particle within the flame front is therefore,

according to Stokes' law:

dv ~S

~=

v,,,. nu = 6irr n (p /pb-l) t - v

at L ot

Setting m = 4/3irr 3p gives a differential equation of the form:

dv

-2- +av - a b t = 0

dt P

9 n ^u

where a = ! and b = (pu/pb-l) .

r pp a

The solution for an initial condition in which both particle and gas are at

rest, gives a particle velocity at t = T of

, b , b -ax

v =bT 1 edL

Paa

Setting n = n = vp = ap , where v is the kinematic viscosity, gives

vv

?* XL= A [i -

where A =

9p

Generated on 2015-01-31 00:36 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

For a typical near-limit condition with Su = 3 cm/sec, a typical density

ratio of p /pu = 3 x 103, a particle diameter of 20ym (r = 10~3 cm) and

'' vg ~ VP

a = 0.55 cm2/sec, one obtains A = 0.02, and hence = 0.02.

vg

Thus the 20ym particle will exit the flame front with a velocity that is

98 pct of the gas velocity, hence it (and particles of smaller size) will

follow the gas flow reasonably well under limit conditions. On the other

hand, a more rapidly propagating flame front, with Su = 30 cm/sec, gives

A = 2.0 and -L-I = 0.8. For the faster flame, the 20ym particle should

22

lag the gas flow quite markedly. Its exit velocity from the flame front would

only be about one-fifth the gas velocity. Thus, although it and the gas enter

the flame front at 30 cm/sec, the gas exits at S - 180 cm/sec whereas the

20pm particle exits at only 60 cm/sec. The particle velocity has thus only

doubled while the gas velocity has increased by a factor of six. This same

magnitude of particle lag would be present even at the limit velocity for a

coarse particle 200ym in diameter. The effect of this particle lag is a sig-

nificant increase in the effective concentration of the dust in the flame

front.

It is important to recognize that both the finite heat capacity effect

discussed earlier, and the particle drag effect considered above, tend to

moderate the -3 particle-size dependence predicted by equation 8. The finite

dP

heat capacity effect limits the quenching effectiveness of very small parti-

cles so that below some critical size in the 1-10 ym range, they can no longer

act as cold walls. The particle drag effect causes the coarser particles to

lag in the accelerating gas flow of the flame front, and as they "pile up"

their effective concentration increases. This increased effective concentra-

tion tends to compensate for the coarser particles' lower surface area. Both

effects tend to reduce the particle-size dependence of the quenching behavior

of inert particles.

Finally, it should be emphasized that these considerations apply only to

inert particles for which Nu -> 1. For such inert particles the boundary layer

thickness is comparable to d , and heat transport is by pure conduction. If,

however, the particle is not inert and can generate volatiles during its pass-

age through the flame front, then its effectiveness may be magnified in two

ways. First if its devolatilization is endothermic, then clearly its effec-

tive heat capacity, C , is magnified, and it would tend to remain colder for

a longer period of time. Equation 11 would then underestimate its t-ratio,

and it would tend to remain as a cold wall down to smaller particle sizes.

Secondly, if the particle generates volatile during its passage through the

flame front, the heat transfer between the gas and the particle is no longer

by pure condition through the boundary layer. Instead, the devolatilizing

gases will mix with the surrounding flame gases on a dimensional scale that

is about an order of magnitude larger than the boundary layer thickness for

Generated on 2015-01-31 00:36 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

an equivalent inert particle. Assuming that the gases are inert or inhibit-

ing, this mixing results in convective cooling (or chemical inhibition) that

is more effective than the pure conduction process used to derive equations 2

and 8.

5There is a molecular analogy to this argument for the addition of gases:

dissociating molecules or molecules with many degrees of molecular motion

would be more effective inerting agents than simpler atomic or diatomic

molecules.

23

CONCLUSIONS

The concept of limit burning velocities, which was previously applied to

the problem of flame quenching by buoyancy (process a) , appears to be applica-

ble to the problem of the thermal quenching of flames by heat losses to exter-

nal contact walls or to internally added inert powders. For wall losses

(process b), the limit velocity is (Su)b = aPe/2r0, and in the presence of

both buoyancy and wall losses, a simple additivity principle appears to apply.

Thus, in the presence of both loss processes, the limit of flame propagation

is reached when the sum of limit velocities, (Su)a + (Su)b , is just equal to

the ideal burning velocity of the mixture, (Su)ideal (which is measured in the

absence of loss processes). The value of the critical Peclet constant, Pe,

for wall loss quenching appears to be sensitive to flame shape and boundary

constraints, but a value in the range Pe - 25 to 60 appears to apply to most

cases of interest.

An analysis of the special problem of downward flame propagation in open

tubes predicts an unusual dependence on tube diameter. This unusual depend-

ence makes it impossible to define a downward flame propagation limit that

would be truly independent of tube size. Downward flame propagation limits

should also display an unusual dependence on boundary constraints, namely,

whether the ignited end is open or closed. These special complications are

absent for either upward or horizontal flame propagation. For those two

directions of flame propagation, "true" limits are observable so long as the

apparatus dimensions are large enough that (S ) (S ) .

For the problem of inert powder addition, a limit velocity is obtainable

of the form: (Su)'b = 0.0083 Pe Cm/ppdp. Measured critical Peclet constants

for heat loss quenching by added inert powders are in the range Pe - 75-175.

Because of finite heat capacity effects, the equation for (Su) 'b is applicable

only for particles above 5-10 ym in diameter. An analysis of the kinematics

of particle motion in an accelerating flame front was also presented. It

indicated that for coarse particles above 50 um in diameter, a significant

particle lag effect exists. This particle lag effect increases the particle's

effective concentration in the flame front. Both the finite heat capacity

effect and the particle lag effect limit the applicability of the above

equation for (S )'b and tend to reduce or moderate the predicted

Generated on 2015-01-31 00:36 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

-j particle-size dependence.

24

REFERENCES

1. Andrews, G. E., and D. Bradley. The Burning Velocity of Methane-Air

Mixtures. Combustion and Flame, v. 19, 1972, pp. 275-288.

2. Coward, H. F., and G. W. Jones. Limits of Flammability of Gases and

Vapors. BuMines Bull. 503, 1952, 155 pp.

3. Edmondson, H., M. P. Heap, and R. Pritchard. Ambient Atmosphere Effects

in Flat-Flame Measurements of Burning Velocity. Combustion and Flame,

v. 14, 1970, p. 195.

4. Egerton, Sir A. C., and S. K. Thabet. Flame Propagation: The Measure-

ment of Burning Velocities of Slow Flames and the Determination of

Limits of Combustion. Proc. Roy. Soc. (London), Ser. A, v. 211, 1952,

p. 445.

5. Friedman, R. The Quenching of Laminar Oxyhydrogen Flames by Solid

Surfaces. Proc. 3d Combustion Symposium, Madison, Wisc., Sept. 7-11,

1948. The Williams and Wilkins Co., Baltimore, Md., 1949, pp. 110-120.

6. Gerstein, M., and W. B. Stine. Analytical Criteria for Flammability

Limits. Proc. 14th Symp. (Internat.) on Combustion, Pennsylvania

State Univ., University Park, Pa., Aug. 20-25, 1972. The Combustion

Institute,Pittsburgh, Pa., 1973, pp. 1109-1118.

7. Gunter, R., and A. Janisch. Measurements of Burning Velocity in a Flat

Flame Front. Combustion and Flame, v. 19, 1972, p. 49.

8. Hertzberg, M. The Theory of Flammability Limits. Natural Convection.

BuMines RI 8127, 1976, 15 pp.

9. . Comment. Proc. 16th Symp. (Internat.) on Combustion,

Massachusetts Inst. Technol., Cambridge, Mass., Aug. 15-20, 1976.

The Combustion Institute, Pittsburgh, Pa., 1977, p. 1404.

10. Hertzberg, M., and K. Cashdollar. The Flammability Limits of Coal

Dust/Air Mixtures. Proc. 1978 Fall Tech. Meeting The Combustion

Institute, Eastern Section, Nov. 29-Dec. 1, 1978. The Combustion

Institute, Pittsburgh, Pa., pp. 35-1 to 35-4.

11. Hertzberg, M., K. Cashdollar, C. Litton, and D. Burgess. The Diffusion

Flame in Free Convection. Buoyancy-Induced Flows, Oscillations,

Radiative Balance, and Large-Scale Limiting Rates. BuMines RI 8263,

1978, 33 pp.

12. Jost, W. Explosion and Combustion Processes in Gases. McGraw-Hill

Generated on 2015-01-31 00:36 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

Book Co., New York, 1946, 93 pp.

13. Levy, A. An Optical Study of Flammability Limits. Proc. Roy. Soc.

(London), Ser. A, v. 283, 1965, p. 134.

25

14. Linnett, J. W., and C. J. S. M. Simpson. Limits of Flammability.

Proc. 6th Symp. (Internat.) on Combustion, Yale University, New Haven,

Conn., Aug. 19-24, 1956. Reinhold Pub. Co., New York, 1957, pp. 20-27.

15. McCaffrey, B. J. Purely Buoyant Diffusion Flames: Some Experimental

Results. Proc. 1978 Fall Tech. Meeting, The Combustion Institute,

Eastern Section, Nov. 29-Dec. 1, 1978. The Combustion Institute,

Pittsburgh, Pa., pp. 8-1-8-4.

16. Potter, A. E., Jr. Progress in Combustion Science and Technology.

Pergamon Press, New York, v. 1, 1960, 145 pp.

17. Potter, A. E., Jr., and A. L. Berlad. The Effect of Fuel Type and

Pressure on Flame Quenching. Proc. 6th Symp. (Internat.) on Combustion,

Yale University, New Haven, Conn., Aug. 19-24, 1956. Reinhold Pub. Co.,

New York, 1957, pp. 27-36.

18. Sapko, M. H., A. L. Furno, and J. M. Kutcha. Flame and Pressure Develop-

ment of Large-Scale CH4-Air-N2 Explosions. Buoyancy Effects and Vent-

ing Requirements. BuMines RI 8176, 1976, 32 pp.

19. Smoot, L. D., and M. D. Horton. Exploratory Studies of Flame and Explo-

sion Quenching. Final Report. V. 1, BuMines Grant No. G0177034,

Mar. 31, 1978, 315 pp. Available at Pittsburgh Mining and Safety

Research Center.

20. Spalding, D. B. A Theory of Inflammability Limits and Flame-Quenching.

Proc. Roy. Soc. (London), Ser. A, v. 240, 1957, p. 83.

21. von Elbe, G., and B. Lewis. Theory of Ignition, Quenching and Stabiliza-

tion of Flames of Nonturbulent Gas Mixtures. Proc. 3d Combustion

Symp., Madison, Wisc., Sept. 7-11, 1948. The Williams and Wilkins

Co., Baltimore, Md., 1949, pp. 68-79.

22. Zabetakis, M. G. Flammability Characteristics of Combustion Gases and

Vapors. BuMines Bull. 627, 1965, 121 pp.

Generated on 2015-01-31 00:37 GMT / http://hdl.handle.net/2027/mdp.39015078466011


Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google

-frU.S. GOVERNMENT PRINTING OFFICE: 1980-603-102/85 INT.-BU.OF MINES,PGH.,PA. 24751

You might also like