You are on page 1of 64

Prog. Energy Combust. Sci. 1987. Voh 13, pp. 97-160. 0360-1285/87 $0.00 +.

50
Printed in Great Britain. All rights reserved. Copyright O 1987 Pergamon Journals Ltd.

RADIATION HEAT TRANSFER IN COMBUSTION SYSTEMS

R. VISKANTA* a n d M . P. M E N G O q t
*School oJ Mechanical Engineering, Pttrdue University, West LaJ~tyette, IN 47907, U.S.A.
tDepartment q/Mechanical Engineeriny, University of Kentucky, Lexington, K Y40506, U.S.A.

Abstract An adequate treatment of thermal radiation heat transfer is essential to a mathematical model
of the combustion process or to a design of a combustion system. This paper reviews the fundamentals of
radiation heat transfer and some recent progress in its modeling in combustion systems. Topics covered
include radiative properties of combustion products and their modeling and methods of solving the
radiative transfer equations. Examples of sample combustion systems in which radiation has been
accounted for in the analysis are presented. In several technologically important, practical combustion
systems coupling of radiation to other modes of heat transfer is discussed. Research needs are identified
and potentially promising research topics are also suggested.

CONTENTS
Nomenclature 98
1. Introduction 98
2. Radiative Transfer 100
2.1. Radiative transfer equation 100
2.2. Conservati_on of radiant energy equation 104
2.3. Turbulence/radiative interaction 104
3. Radiative Properties of Combustion Products 106
3.1. Radiative properties of combustion gases 107
3.1.1. Narrow-band models 107
3.1.2. Wide-band models 107
3.1.3. Total absorptivity emissivity models 109
3.1.4. Absorption and emission coefficients 109
3.l.5. Effect of absorption coefficient on the radiative heat flux predictions 111
3.2. Radiative properties of polydispersions 113
3.2.1. Types and shapes of polydispersions 114
3.2.2. Prediction methods of the particle radiative properties 115
3.2.3. Simplified approaches 116
3.2.4. Scattering phase function 119
3.3. Total properties 121
4. Solution Methods 122
4.1. Exact models 122
4.2. Statistical methods 123
4.3. Zonal method 123
4.4. Flux methods 124
4.4.1. Multiflux models 125
4.4.2. Moment methods 126
4.4.3. Spherical harmonics approximation 127
4.4.4. Discrete ordinates approximation 128
4.4.5. Hybrid and other methods 129
4.5. Comparison of methods 130
5. Applications to Simple Combustion Systems 133
5.1. Single-droplet and solid-particle combustion 133
5.2. Contribution of radiation to flame wall-quenching of condensed fuels 133
5.3. Effect of radiation on one-dimensional char flames 134
5.4. Radiation in a combusting boundary layer along a vertical wall 135
5.5. Interaction of convection-radiation in a laminar diffusion flame 136
5.6. Effect of radiation on a planar, two-dimensional turbulent-jet diffusion flame 138
5.7. Radiation from flames 139
5.8. Combustion and radiation heat transfer in a porous medium 140
6. Applications to Combustion Systems 141
6.1. Industrial furnaces 142
6.1.1. Stirred vessel model 143
6.1.2. Plug flow model 145
6.1.3. Multi-dimensional models 145
6.2. Coal-fired furnaces 146
6.3. Gas turbine combustors 149
6.4. Internal combustion engines 151
6.5. Fires as combustion systems 151
7. Concluding Remarks 153
Acknowledgements 154
References 154
apses 1 3 : z - x 97
98 R. VISKANTAand M. P. MENGf3t~

NOMENCLATURE direction cosine, Eq. 12.8)


p density (kg/m 3)
A a r e a [m 2)
a Stefan-Boltzmann constant; scattering CO-
B mass transfer number, (Q Y~o/vowo-h.)/L efficient (m- 1)
C concentration
T beam transmittance, Eq. (2.18)
D diameter of particles (am) or burner exit r optical depth, i,y
diameter (m)
~P scattering angle, Eq. 2.7)
D, dimensionless heat of combustion.
scattering phase function
(2 V,,/v ,, W,,h,,.
05 azimuthal angle
Eh blackbody emitted flux defined by aT4(W/m z)
solid angle
E,, exponential integral function,
t,) single scattering albedo, a/fl
E,,tx)= So~" - 2exp( -x/la)d/d
f size distribution, Eq. (3.11) or phase function
coefficient, Eq. (3.25)
Subscripts
J/0) dimensionless stream function at the surhce
./v volume fraction (m3/m 3) b refers to blackbody
g phase function coefficient, Eq. (3.25) e refers to effective mean
h enthalpy or Planck'sconstant i refers to Planck's internal mean
I radiation intensity(W/m 2-sr) m refers to mean values
J radiative flux in radial direction (W/m 2) n refers to narrow-band model
K radiative flux in axial direction (W/m 2) o oxygen
k thermal conductivity (W/mK) or imaginary P refers to Planck's mean
part of the complex index of refraction r refers to spatial coordinates or radiation
Ko Konokov number II' refers to wall conditions, fuel surface or wide-
L radiative flux in angular direction (W/m z) or band model
effective latent heat of pyrolysis 2 refers to wavelength dependent properties
L,, mean beam length (m) V refers to frequency dependent properties
Mp pyrolysis rate
Nn radiation-conduction parameter, k h / o T 3
N~ conduction-gaseous radiation parameter, Stlperscripts
k~x/aT 3
N2 conduction-ambient radiation parameter refers to incoming radiation beam
defined as limy~ (k~o/aT3Xu~/xv) 1/2 refers to turbulent mean properties
h complex index of refraction (= n - i k )
n real part of the complex index of refraction
P pressure or probability density function I. I N T R O D U C T I O N
PN N-th order spherical harmonics approxi-
mation Expenditures on fossil energy by individuals,
Q Mie efficiency factor or energy released by commerce, transportation and industry in an in-
combustion of v moles of gas phase fuel
Qv heat of reaction per unit mass of oxygen dustrialized country account for a significant fraction
Re burner Reynolds number of the country's G N P . Improved understanding of
q heat flux (W/m 2) combustion systems which use fossil fuels such as
r mass consumption number natural gas, oil and coal may result in improved
S source function
energy efficiency. The potential improvement in the
S~ N-th order discrete ordinates approximation
s coordinate along the direction of propa- thermal performance of such systems could make a
gation of radiation or stoichiometric ratio, significant impact on the country's economy. This
vs, W / v o W,, provides the motivation and economic incentive for
St Stanton number research and development in combustion technology.
T temperature (K)
T,o surface temperature of load (sink) (K) An important goal, then, is to develop computational
v velocity (m/sec) models which could be used for the design and
V volume (m 3) optimization of more cost effective and environ-
W~ molecular weight of species i mentally friendly combustion systems with improved
x size parameter, riD~2
performance.
y~ mass fraction of species i
Combustion is one of the most difficult processes
to model mathematically since it generally involves
Greek letters the simultaneous processes of three-dimensional two-
absorptivity phase fluid dynamics, turbulent mixing, fuel evapor-
extinction coefficent (m - ~) ation, radiative and convective heat transfer, and
6 Dirac delta function
chemical kinetics. In order to design combustion
emissivity
systems based on fundamental principles, compre-
emission coefficient, Eq. (2.4); direction co-
sine, Eq. (2.8); dimensionless coordinate hensive models incorporating all of these factors are
defined as x/g~I~(u/tt)dy required. State-of-the-art reviews of modeling some
zenith angle; normalized temperature combustion systems have been p r e p a r e d ? - 7 Signif-
absorption coefficient (m - ~) icant progress has been made in detailed modeling of
wavelength of radiation (/tm)
combustion systems, but major problems such as
direction cosine, Eq. (2.8)
frequency of radiation; kinematic viscosity: turbulence in reactive flows, particle formation and
stoichiometric coefficient others remain to be solved.
Radiation heat transfer 99

An adequate treatment of thermal radiation is ~= - k V T + ~ k + ~ n i h i V j + ~ a _ ,. (1.2)


essential to develop a mathematical model of the J
combustion system. The level of detail required for In Eq. (1.1) p, pc, ~ and P are the total mass, energy
radiative transfer depends on whether one is inter- density, fluid velocity, and pressure, respectively. The
ested in determining the instantaneous spectral local {n;} and {V~} are the number density and diffusion
radiative flux, flame structure, scalar properties of the velocities of the individual chemical species, and ,~"
flame, formation of flame-generated particles (largely is the radiation heat flux vector. The first, second,
soot), local radiative flux and its divergence or the third and fourth terms in Eq. (1.2) account for
temperature distribution. For example, when the molecular conduction, radiation, interdiffusion and
model is used to predict pollutant concentrations, diffusion-thermo contributions, respectively, to the
accurate temperatures are especially important since heat flux vector. In Eq. (1.1) S is the local volumetric
the chemical kinetics involved are extremely temper- heat source/sink from other processes, if any. When
ature dependent. radiation heat transfer needs to be accounted for in
The fraction of the total heat transfer due to the energy equation, it is preferable to use temper-
radiation grows with combustor size, attaining ature as the dependent variable rather than the
prominence for gaseous firing at characteristic com- stagnation enthalpy. The divergence of the radiative
bustion lengths of about 1 m. Radiation heat trans- flux vector, V..~-~', can be obtained from the radiant
fer, then, plays a dominant role in most industrial energy equation.
furnaces. Unfortunately, it is governed by a complex The purpose of this paper is to acquaint the reader
integrodifferential equation which is time consuming with the basic principles and methods related to
to solve. Economic measures are a necessity, even at modeling radiation heat transfer in combustion
the loss of some accuracy. systems. The importance of radiative transfer in coal
In a combustion chamber, radiation heat transfer combustion, 3 pulverized coal-fired boilers,'* indus-
from the flame and combustion products to the trial furnaces, 5 gas turbine combustors 6 and fires 7
surroundings walls can be predicted if the radiative has been recognized for some time. Radiative
properties and temperature distributions in the transfer in some of these systems has received
medium and on the walls are available. Usually, considerable research attention and a high degree of
however, temperature itself is an unknown para- organization has been attained.
meter, and as a result of this, the total energy and The paper is organized to give a systematic and
radiant energy conservation equations are coupled, easy-to-follow approach to the major building blocks
as in many heat transfer applications. Solution of the of radiative transfer in combustion systems. Section
thermal energy equation can be obtained if several 2 of the paper introduces the fundamentals of
other physical and chemical processes can be radiation heat transfer, and Section 3 discusses the
modeled. The major processes which need to be radiative properties of gases and particles encoun-
considered in a combustion system in addition to tered in combustion systems. These two sections
radiation include? (i) chemical kinetics, (ii) thermo- provide the background necessary for understanding
chemistry, (iii) molecular diffusion, (iv) laminar and the specific techniques for solving the radiative
turbulent fluid dynamics, (v) nucleation, (vi) phase transfer equation discussed in Section 4. Examples of
transitions, such as evaporation and condensation simple combustion systems in which radiative trans-
and (vii) surface effects. Since the physical and fer has been accounted for are discussed in Section 5.
chemical processes occurring in combustion cham- Section 6 reviews modeling of radiation heat transfer
bers are very complicated and cannot be modeled on in practical combustion systems and deals with
the microscale, there is a need for physical models to coupling of radiation to other transport processes in
simulate these processes. Each of these models needs system models.
an extensive and separate treatment, which is outside There exists a very large body of literature relevant
the scope of this work. The interested reader is to radiation heat transfer in combustion systems, and
referred to more specialized publications. ~- v it is not possible to cover it thoroughly. Emphasis in
In nonrelativistic problems of an engineering the paper is on fundamentals and applications to
nature, radiation does not contribute any terms to simple systems. Reference is made to the original
the conservation of mass, momentum and species publications for a more complete discussion. A
conservation. The classical conservation of energy review process is a rather arbitrary activity, because
equation Ls'9 is modified by a contribution which of the decision the authors have to make on what to
accounts for radiation heat transfer. This equation include, what to omit, and where to start and end.
can be written as This article is no exception, and it reflects the
authors' biases. Because of the broad range of topics
~pe covered, details can not be included, and no claim is
= -V.pe~-V.P-~-V.~+S (1.1)
Ft made as to the completeness of the review. In these
days of many journals and other publications, it is
possible that relevant work may have been inad-
where the heat flux vector, ~, is defined as vertently overlooked.
100 R. MISKANTAand M. P. MENG0q

2. RADIATIVE TRANSFER magnetic field, and Osborn and Klevans ~4 have


refined and generalized their work. The Eulerian
2.1. Radiative Transfer Equation
point of view is adopted here and the traditional
Two theories have been developed for the study of intuitive derivation of the RTE found in the
the propagation and interaction of electromagnetic radiative transfer literature 15 - 20 is given.
radiation with matter, namely, the classical electro- Rather than presenting the most general deriv-
magnetic wave theory and the radiative transfer ation of the RTE, certain constraints which help to
theory. The theories were developed independently avoid complications that obscure the physical signifi-
and there is no similarity in their basic formulations. cance of the phenomenon are imposed in this
Conceptually, they are completely distinct; however, discussion. The treatment presented here constitutes
both theories describe the same physical phen- a reasonable compromise between the generality
omenon. The classical electromagnetic theory has needed for engineering applications and clarity of the
approached the study of propagation and interaction development. The idealizing assumptions and con-
of matter with radiation from the microscopic point straints imposed are: (1) the discussion is restricted to
of view and the radiative transfer theory from the a continuous, homogeneous and isotropic absorbing-
macroscopic (or phenomenological) point of view. emitting-scattering medium at rest, (2) the state of
The study of the detailed interaction of electro- polarization is neglected, and (3) the medium is
magnetic radiation with matter on the microscopic considered to be in local thermodynamic equilibrium
level from both the classical and quantum mechanics (LTE).
point of view yields the interaction cross-sections of The RTE is based on application of an energy
the particles making up the matter. This fundamental balance on an elementary volume taken along the
approach predicts the macroscopic properties of the direction of a pencil of rays and confined within an
media, and these properties appear as coefficients in elementary solid angle. The detailed mechanism of
the radiative transfer equation. the interaction processes involving particles and the
The quantitative study, on the phenomenologicai field of radiation is not considered here. On the
level, of the interaction of radiation with matter that phenomenological level only the transformation
absorbs, emits, and scatters radiant energy is the suffered by the radiation field passing through a
concern of the radiative transfer theory. The theory participating medium is examined. The derivation
ignores the wave nature of radiation and visualizes it accounts mathematically for the rate of change of
in terms of light rays of photons. These are concepts radiation intensity along the path in terms of
of geometrical optics. The geometrical optics theory physical processes of absorption, emission, and
is the study of electromagnetism in the limiting case scattering.
of extremely small wavelengths or of high frequency. Consider a cylindrical volume element, Fig. 1, of
The detailed mechanism of the interaction process cross-section dA and length ds in an absorbing,
involving atoms or molecules and the radiation field emitting, and scattering medium characterized by the
is not considered. Only the macroscopic problem spectral absorption coefficient xv, scattering coef-
consisting of the transformation suffered by the field ficient try and true emission coefficient r/v. The axis of
of radiation passing through a medium is examined. the cylinder is in the direction of the unit vector ~,
Thus, there is a considerable simplification over the i.e. ds is measured along ~. The spectral intensity of
electromagnetic wave theory. radia.tion (spectral radiance) in the ~-direction
The radiative transfer equation (RTE) forms the incident normally on one end of the cylinder is Iv
basis for quantitative study of the transfer of radiant and the intensity of radiation emerging, through the
energy in a partici, pating medium. The equation is a second end in the same direction is Iv + dlv. Here, v is
mathematical statement of the conservation principle the frequency and is related to the wavelength 2 by
applied to a monochromatic pencil (bundle) of v = c/2, where c is the velocity of radiation.
radiation and can be derived from many viewpoints.
Some authors lA~ have derived the radiative
transfer equation from the Boltzmann equation of
the molecular theory of gases by adopting the
corpuscular ' picture of radiation and recognizing
close analogy between molecules and photons.
Preisendorfer ~2 has presented a development prim-
arily from the standpoint of geometrical optics by
starting from a set of physically motivated axioms
from which the features of radiative transfer were
deduced. Several papers have also considered the
derivation of the equation from quantum mechanics.
Harris and Simon ~3 used the Liouville equation to
consider coherent radiation from a plasma by a FIG. 1. Coordinates for derivation of the radiative mmsfer
statistical treatment of both plasma particles and the equation.
Radiation heat transfer 101

It follows from the definition of the spectral Since in this case the sum of probability over all
intensity I, that radiant energy incident normally on directions must equal unity, we must have
the infinitesimally small cross-section dA during time
interval dr, in frequency range dv and within the 1 1
elementary solid angle dD about the direction of the 4rt n'=4, a=4,
unit vector ~ is
1
l ,dAdfldvdt. =-- J" ~,(W)dfl= 1. (2.5)
4n n=4,
The emerging radiant energy at the other face of the
This implies that for coherent scattering the spectral
cylinder in the same direction equals
phase function is normalized to unity. The scattering
angle W, i.e. the angle between g' and g can be
(1 ~+ dl ,)dAdDdvdt.
expressed as
The net gain of radiant energy, i.e. the difference
cos W = cos0cos0' + sin0sin0'cos(~ - ~b') (2.6)
between energy crossing the two faces of the cylinder,
is then given by
or

(I ~+ dl ~- dl,)dAdf~dvdt = dl,dAdfldvdt. (2.1)


cos W = ~'+ qq' + I#~' (2.7)
The loss of energy from this pencil of rays due to
where
absorption and scattering in the cylinder is
~=sin0cos4~, q=sin0sin~b, /~=cos0 (2.8)
( K, + ~r,)l ,dsdA d~dvdt. (2.2)
are the direction cosines in any orthagonal co-
The emission by the matter inside the cylindrical
ordinate system. Reference to Fig. 1 shows that
volume element dV, in the time interval dt, in the
g'(0',~b') represents the incoming direction of the
frequency range dr, confined in the solid angle dD
pencil of rays, and g(0,~b), is the direction of the
about the direction g equals
pencil after scattering.
Equating the change of energy in the cylindrical
q~d V df~dvdt. (2.3)
volume element to the net gain or loss of energy
along the traversal path of the cylinder in terms of
If Kirchhoff's law is valid, the emission coefficient q~
the processes of attenuation, emission and in-
can be expressed as
scattering yields
rl~ = x~n~lh~ (2.4)
d/,dA dfldvdt = - (x, + a ,)l ,dsdA dDdv dt
where lb, is Planck's spectral blackbody intensity of
II ~ ,, ~,
radiation, and n, is the spectral index of refraction of +q,dVdlldvdt + a , d s ~ A," S q~,(s ---~s;v---w)
the medium. The increase in energy of the pencil of fl' = 4It

rays (~,df~) due to in-scattering of radiation by the


matter into the elementary cylindrical volume from
all possible directions ~' is x l(-~')dt)'dv'] dAdfldv dt. (2.9)

,(s ~ s ; v ~ v )
&v' fl' ='l-x
Dividing this equation by dAdsdf~dvdt and recalling
that the distance ds traversed by the pencil of rays is
cdt, where c is the velocity of light in the medium,
x l(-~')dD'dv']dAd~dvdt. yields the equation of transfer in a Lagrangian
coordinate system
In this expression the phase function 1 dlv
t~,(g'---,g;v'---~v)df~'dv'/4n represents the probability . . . . (x, +tr,)lv + q ,
that radiation of frequency v' propagating in the c dt
direction g' and confined within the solid angle dfg
O" v
is scattered through the angle (g,g) into the solid @,(s" -*s;v-*v)l,,(s )dD' dv.' (2.10)
+U.I I ~' "'
angle dD and the frequency interval dr. This proba- Av" 11" = 4 x
bility is determined by the scattering mechanism.
For coherent scattering the phase function is inde- Clearly, the left-hand side of this integrodifferential
pendent of frequency v' and reduces to ~v(~"--*g). equation represents the net change in Iv per unit
102 R. V1SKANTAand M. P. MENGOt;

length along the path ds=cdt. Equation (2.10) is a Z,


statement of the conservation of energy principle for (a) T
a monochromatic pencil of radiation (in the dir-
ection g) and is generally called the "radiative
transfer equation" (RTE). In some literature, the
steady-state form of this equation is called Bouguer's
law probably due to the fact that the constitutive
(Bouguer's) law enters as the first term on the right-
hand side of Eq. (2.10).
/ f /
The substantial derivative d/dt refers to the rate of
change of spectral intensity as seen by an observer
propagating along with the velocity of radiation
(Lagrangian coordinates). In terms of a coordinate
system fixed in space (Eulerian coordinates), RTE (b) A s
may be written as

1 dl, 1 01,
~-(V.-~)I,=fl,(S,-I,) (2.11)
c dt c 0t

where the source function S, represents the sum of


emitted and in-scattered radiation and is defined as
radiant energy leaving an element of volume of
matter in the direction (g,df]) per unit volume, per
unit solid angle, per unit frequency, and per unit (c)
ti me,

sv = (,lv//L) + (a,//LX1/4~)

O,(s s;v v)/(s )df~dv. (2.12)


Av' fl' = 4ft
:.,y

It is evident on inspection of Eq. (2.11) that there is


no net rate of change of Iv at a point, if and only if,
FIG. 2. Coordinates for Cartesian (a), cylindrical (b) and
lv=Sv. If I~>S, then dIv/dt<O so that as t spherical (c) systems.
increases, I , is decreasing toward S , and if I , < S ,
then dl,/dt > 0, so that I~ is increasing toward S,.
Equation (2.11) can be written explicitly using the
analytical forms of 07. g)l, which are given in Table remains invariant under the rotation about the z-
1. The direction cosines ~, q, and/~ are defined by Eq. direction. This allows us to combine the two
(2.8), and they are functions of angular variables 0 azimuthal angles, namely ~b and ~b,, to obtain a single
and ~b. Usually, the polar and aximuthal angles for azimuthal angle. If, one writes ~b'=q~-~b,, then
space variables are also designated by 0 and ~b. To 07. g)l, for an axisymmetric cylindrical system can
avoid confusion and to be able to use the con- be given as
ventional nomenclature at the same time, we choose
to use subscript r for space variables 0, and ~b, when 7 01, rl 01, 01, (2.13)
appropriate (see Table I). ~ ) 1 , = ~ - - ; o , ~u ez"
In Fig. 2, three orthogonal coordinate systems and
the corresponding nomenclature are shown, where
the spherical coordinate system for angular variation In a spherically symmetric system, the radiation
of intensity is superimposed on either a rectangular, intensity depends on only two parameters, i.e. the
cylindrical or spherical coordinate system for spatial radial distance r measured from the origin and the
variables. In general the intensity is a function of direction cosine/~ of the angle between the direction
three spatial coordinates, two angles and time; of of the radiation beam and the radius vector it . The
course, the seventh independent variable required to analytical expression for 07. g)l, corresponding to a
define the radiation intensity is the wavelength or the spherically symmetric system is the last expression of
frequency of radiation. Table 1, which can be simplified further to read
For most practical calculations, it is possible to
assume cylindrical or spherical symmetry. The cylin- 07" ~)1 __u 0/,+1 - u 2 0t, (2.14)
drical symmetry requires that the radiation intensity
Radiation heat transfer 103

A number of assumptions have been made in the


derivation of the RTE, and for the sake of complete-
ness it is desirable to discuss them briefly. The first
assumption concerning the restriction that the par-
ticipating media be continuous, homogeneous and
isotropic has been relaxed by Preisendorfer. 12 Al-
though the assumption of a medium at rest is open to
criticism on physical grounds, this approximation
correctly describes all engineering problems where
the fluid velocity is much smaller than the velocity of
light. The absorption and scattering coefficients are
calculated or measured in a laboratory reference
system in which the local macroscopic velocity of
matter is zero, and because of this xv, av and T are
independent of g. It has, however, been shown that
FiG. 3. Coordinates for radiative transfer along a line-of-
in any frame of reference Iv satisfies the same
sight.
equation of radiative transfer. 22 The intensity Iv
changes at points along the path, where the index of
refraction n, changes continuously or discontin- function of time indirectly through the source
uously. Such changes can be systematically accounted function if qv is time dependent [see Eq. (2.12)].
for by simply adopting a new function l,/n~ rather Suppose that at some point on the boundary of
than Iv. 12 Hence, there is no need to include the matter So, as shown in Fig. 3, the spectral intensity Iv
index of refraction explicitly in the transfer equation. is known
The second assumption concerning the fact that
neglect of polarization is not generally valid is well lv(s)=l,(so)=lo~ at s = s o. (2.16)
recognized, and it is clear that polarization must be
accounted for in any rigorous treatment of radiative The integral form of the equation of transfer may be
transfer when scattering is present. The radiative derived from the integrodifferential equation by
transfer theory has been extended to include the imagining the latter to be an ordinary differential
phenomenon of polarization of radiation. 12'15 It is
equation in the unknown Iv and with S, a known
also well recognized that the third assumption for the source function. The integrating factor for this
medium to be in LTE may be invalid under the differential equation is exp (Sflvds) and the integral of
conditions where densities and optical thicknesses Eq. (2.15) with the boundary condition Eq. (2.16) may
become small, scattering becomes an important be written as
mechanism, rapid time variations occur or large
temperature gradients are in evidence? 2 Therefore,
I v(s) = I o ffv(S,So)
before making the LTE assumption, the conditions
for a given physical system should be carefully
i t !
examined. + f f s (s)T (s s )fl (s)ds
v v ~ v
(2.17)
The radiative transfer equation, Eq. (2.11), is an
integrodifferentiai equation, and because of this it is
very difficult to solve exactly for multidimensional where s' is a dummy variable of integration and
geometries. Therefore, some simplifications of this T(s,s') is the beam transmittance of an arbitrary path
equation are necessary. A close look at the source from s' to s along the direction
term given in Eq. (2.12) reveals that the in-scattering
term (the second term of the right hand side) yields
the integral nature of the RTE. If scattering is
negligible in the medium, then the Eq. (2.11) will be a
T~(s,s)=exp
if':/ ,]-
=
flv(Od . (2.18)

linear differential equation, which is much easier to The concept of the beam transmittance can be made
solve than the linear integrodifferential equation. clearer by the following interpretation. If lov repre-
A formal solution of the quasi-steady state RTE, sents the intensity of radiation in some direction g at
Eq. (2.11), can readily be written. Consider a pencil of some initial point So and Iv(s) is the intensity of the
radiation in the direction g (Fig. 3). If the coordinate transmitted radiation at point s in the same direction
s is laid in the direction g, the quasi-steady RTE is over the path from the initial to the terminal points,
given by then the two intensities are related by

(V- ~)1,= ~g*= f l , ( S v - I O (2.15) l,(s)= T~(s,so)lov. (2.19)

where the direction of the pencil of rays is under- Thus, the beam transmittance represents the fraction
stood to be g. The intensity, however, may be a of the initial intensity which is transmitted without
104 R. VISKANTAand M. P. MENGO~3

emission or scattering contributions to the intensity energy from an element of matter per unit of volume
along the path length. Equation (2.17) gives the and per unit of frequency. The term 4nqv(= 4nxJb,)
spectral intensity of radiation at a point and in a represents the local rate of emission, and x,ff,
given direction. Its physical meaning can be more represents the local rate of absorption of radiation
readily interpreted by referring to Fig. 3. It shows per unit of volume. The meaning of the terms can be
that Iv(s) is a sum of two contributions: (1) the further clarified when we note that 4nlbv is the
transmitted intensity, and (2) the path intensity. The product of the spectral radiant energy density of a
first term on the right hand side of Eq. (2.17) is the black body at the local temperature, times the local
contribution to Iv due to the initial intensity at point velocity of light c, while c~v is related to the local
So in the direction of propagation of the radiation g, radiant energy density of space as defined by Eq.
attenuated by the factor Tv(s,s0) to account for (2.21b). In deriving Eq. (2.20) the scattering terms
absorption, scattering and induced emission in the have canceled out. This just confirms the physical
intervening matter. The second term results from fact that scattered energy is not stored and should
both emission and scattering from elements of the not appear in the conservation of radiant energy
matter at all interior points, each elementary contri- equation. Integration of Eq. (2.20) over the entire
bution being attenuated by the factor Tv(s,s') while spectrum results in the conservation equation of total
the rest is absorbed and scattered along the path. radiant energy
These elementary contributions are integrated over
oo
all the elements between the boundary of the body s o
and the point s. 0d + V . , ~ = f 1%[4nlbv(T)-f~v]dv. (2.22)
~t ~o
We note that the integral form, Eq. (2.17), of the
radiative transfer equation is referred to as "'formal For reasons that were explained in a previous
solution" in the sense that I v is expressed in terms of subsection, the time rate of change of radiant energy
integrals that can be evaluated only if the state of the density q/ can be neglected. Note that there is no
matter and the radiation field, i.e. Sv is known. This convective term in Eq. (2.22), since radiation propa-
does not mean that the equation of transfer in a gates inependently of the local material velocity. The
participating medium has been solved. It is clear that equation describing the local change of radiant
if the source function depends on the intensity Iv in energy density must be modified in the relativistic
some specified way, then one can convert Eq. (2.17) treatment of electromagnetic radiation.l'23 How-
into an integral equation for Iv. ~5 However, before ever, the additional terms which arise in the
we do this it is desirable to derive the conservation of conservation of radiant energy equation can gener-
radiant energy equation. ally be ignored in engineering applications.
It is worth noting that the spectral dependence of
radiative properties is denoted either by subscript v
2.2. Conservation of Radiant Energy Equation (frequency) or ~. (wavelength). If the matter through
Integration of the RTE, Eq. (2.11), over all which radiation is propagating is not homogeneous
directions results in and uniform, then the index of refraction, and, as a
result of this, the wavelength and speed of light
would be different at different locations in the
t3ail* + V',~rv=x,[4nlbv(T)--f~v] (2.20) medium, whereas the frequency remains constant
dt
everywhere. Therefore, the frequency is a more
fundamental measure than the wavelength of radi-
where the spectral radiant energy density q/v, the
ation, and because of this, here, the spectral depend-
irradiance aJv and the radiation flux vector "~'v are
ence is denoted by v. It is also useful to remember the
defined as
identity, -lvdv = lad2, between frequency and wave-
length based definitions of radiation intensity.
q/=lf l,df2 (2.21a)
C JQ=4~
2.3. Turbulence~Radiation Interaction
c~v= S lvdfl=cq/v (2.21b) Interaction of convection and radiation has been
12=4~ recognized for some time, but the fact that turbulence
can influence radiative transfer and vice versa has
~,= S 1,~dn (2.21c) been recognized more recently. The first attempt at
D=4x combined analysis of the equations for the mean-
square fluctuations of the velocity and temperature
respectively. The physical meaning of Eq. (2.20) is fields with the radiation field is due to Townsend. 24
clear. It is the conservation equation of spectral Applications in which radiation/turbulence inter-
radiant energy. The right-hand-side of Eq. (2.20) action may affect flow and heat transfer include
represents the net rate of loss or gain of radiant industrial furnaces, gas combustors, flames and
Radiation heat transfer 105

fires.25- 3o Most studies concerned with modeling of Turbulence can influence radiative transfer through
radiative transfer in combustion chambers and fluctuations in temperature and radiating species
furnaces have ignored the turbulence/radiation inter- concentrations which, in turn, influence Planck's
action. 3's An up-to-date discussion of the interaction function lba(T) and the special absorption and
in flames is available 3~ and need not be repeated scattering coefficients. The fluctuations of the Planck
here. Suffice it to mention that the interactions and function and the spectral absorption and scattering
coupled effects are more important for luminous than coefficients can be given in terms of the temperature
for nonluminous flames. Little is known concerning and species fluctuations by means of Taylor series
temporal aspects of radiative transfer in turbulent expansions about the values evaluated at the mean
flames as these effects have not been studied properties. Evaluation of the instantaneous intensity
extensively. of radiation in terms of the mean and fluctuating

TABLE1. Analytical forms of (V.~)l in common orthogonal geometries 2~


(= sin0cos~, tI = sin0sin~b,p = cos0)

Space Direction
Geometry wmables cosincs (V. ~)1

FI ?1 ?1
Rectangular .\.y.: ~..q ,u ,~-- + ll--E- + p - -
.\" ( y ,r -

?1 ;I
-\'.Y ~,q ~ - + 'I S -
~.\- ~y

;I
- It P-5
(E

F(rl) tl FI ;1 1 ?UII)
Cylindrical r,dp~.: ~..q.fl - -- -~ - -- + p,
r ;r r?#), ~- r ?oh

;trl) ;/ I ?(ql)
r.- ;~.q,ll ---- + It
r ?r ;z r ;(~

?(rl) ~1 ?l 1 ?Off)
r.4~, ~.ll --- +
r ;r ri'q~, r ;b

;~ ?(rl) 1 ?[ql)
r ~,l I
r ?r r ,"~

It 70"21) g. ?(sinOfl)
Sphcrical r.O,.q'), '.q.p ------ +
r 2 Fr rsinO~ i'O,

q 21 I ?[{I-p2)l]
rsin0, ;q~, r ?/I

cotO ?(ql )
r ;q~

p ~(r21) ~ ;.(sin(lfl)
- - _ _ q - _ _
r.O, ~dl.l I
r 2 ~r rsin0, ;0,

1 ?[tl-p2)l] cotO, flql)


+
r ;p r ;~

p ?(r21) I ;[(I-/~2)I]
- - -b
r2 2r r ?I~
106 R. VISKANTAand M. P. MENGf3~:

properties and the time-averaging is straight-forward thin. a We further assume that the properties of the
but tedious. 27 If the absorption coefficient can be fluctuating eddies are statistically independent, and
expressed as this implies that there is no correlation between the
temperature and concentration within each eddy.
xa(s,t) = ~,kai Ci(s,t) (2.23) Under these conditions radiation is transmitted
i through an eddy with little change so that the
radiance at a local point is affected little by the local
the turbulent fluctuations in the absorption coef- fluctuation of xx. Hence, the time-average RTE can
ficient can be related to those in the concentrations be approximated as a
Ci of the radiating species. The precise evaluation of
the time-average would utilize the joint probability (V' g)la = - xala + qa. (2.26)
density function P(T,Ci,s) of the temperature and
species concentrations for all points s along the line Following a similar argument, the spectral radiant
of sight g in Eq. (2.17). Unfortunately, that infor- energy Eq. (2.20) can be expressed as
mation is not available. Those properties of the flow
field that are available are the mean temperature T,, V" "#'a = - x-a~a + 4r~Oa. (2.27)
species concentrations C , and the second order
correlations, T '2, C~T'. To illustrate the nature of the Information necessary to solve Eq. (2.25) for the
problem we restrict ourselves to a single radiating time-averaged spectral radiance I~ is not available,
species and neglect scattering. Applying Reynolds' and the integration of Eq. (2.24) along the line-of-
averaging techniques to Eq. (2.17) but omitting the sight is too time consuming. Some clever way of
details, one can obtain 30 ensemble averaging the radiance or developing
correlation coefficients for time-averaged quantities
s s
will be required to enable solution of the integral or
ia(s) = loaexp[ - kaIC(s')ds']exp[ -- kaSC'(s')ds']
differential forms of the RTE in turbulently fluctu-
0 0
ating media. The significance of the turbulence/
s s s radiation interaction will be assessed later.
+ I~(s')exp[ - k~ICds"] {exp[ - kaIC'ds"]
0 s' s'

s 3. R A D I A T I V E PROPERTIES OF COMBUSTION PRODUCTS


+(qffr/])exp[-kaIC'ds"] }ds'. (2.24)
s' The accuracy of radiative transfer predictions in
combustion systems cannot be better than the
This equation can be written in a more useful form in accuracy of the radiative properties of the combus-
terms of the two-point correlation coefficients. 28 The tion products used in the analysis. These products
representation of the random concentration and usually consist of combustion gases such as water
temperature by Gaussian variables is convenient, but vapor, carbon dioxide, carbon monoxide, sulfur
it must be noted that they encompass unrealistic dioxide, and nitrous oxide, and particles, like soot,
negative values of the variables whose probability fly-ash, pulverized-coal, char or fuel droplets. Before
must be kept small in proportion. Comparison of attempting to tackle radiation heat transfer in
Eqs (2.17) and (2.24) reveals that consideration of practical combustion systems, it is necessary to know
turbulence (i.e. time-averaging) would greatly in- the radiative properties of the combustion products.
crease the computational effort of an already difficult Considering the diversity of the products and the
problem. probability of having all or some of these in any
An alternative to time-averaging the spectral volume element of the system, it can easily be
radiance would be to time-average the quasi-steady perceived that the prediction of radiative properties
form of RTE, Eq. (2.15), and the radiant energy in combustion systems is not an easy task. The
equation, Eq. (2.20), at the start. Time-averaging of wavelength dependence of these properties and
Eq. (2.15) results in uncertainties about the volume fractions and size and
shape distribution of particles cause additional
0 7 " g ) ~ = - x a l a + rlx. (2.25) complications.
In order to present a systematic methodology for
The difficulty with this equation is the evaluation of the prediction the radiative properties of combustion
the coupled correlation xala because instantaneous products, the discussion in this section will be
I~ is expressed in terms of an integration along the divided into several subsections in which the re-
path as indicated in Eq. (2.17). To simplify the lations for obtaining the properties of the combus-
absorption coefficient-radiance correlation x~la we tion gases and different particles are discussed and
can assume that the individual eddies are homogen- the simplifications are introduced. Afterwards, some
eous and that the radiating gas of a typical size eddy relations will be given to employ these expressions as
(i.e. based on macroscale of turbulence) is optically building blocks to determine the radiative properties
Radiation heat transfer 107

of the mixture of combustion products. Note that charged particles, then the Stark profile yields a more
usually the level of simplification for the properties is accurate representation of the spectral line radi-
to be determined by the user, and it should be ation. 33 Note that it is also possible to superpose
consistent with the level of sophistication of the these line profiles to incorporate the effects of
radiative transfer and total heat transfer models. different physical conditions on the line radi-
Also, the relations for the radiative properties of ation.33. 3'*
individual constituents should be compatible with There are basically two ~different line arrangements
each other as well as with the radiative transfer for narrow band models used extensively in the
models. literature. The Elsasser or regular model assumes that
the lines are of uniform intensity and are equally
spaced. The Goody or statistical model postulates a
3.1. Radiative Properties o f Combustion Gases
random exponential line intensity distribution and a
Every combustion process produces combustion random line position selected from a uniform
gases, such as water vapor, carbon dioxide, carbon probability distribution. For practical engineering
monoxide, and others. The partial pressures of these calculations both of these models yield reasonably
gases in the combustion products are determined by accurate results. Usually there is less than 8~o
the type of the fuel used and the conditions of the discrepancy between the predictions of these two
combustion environment, such as fuel-air ratio, total models. 35 A detailed discussion of the narrow band
pressure and ambient temperature. These gases do models has been given by Ludwig et al. 34 and in the
not scatter radiation significantly, but they are strong review articles by Tien 33 and Edwards. 3S
selective absorbers and emitters of radiant energy. Narrow-band model predictions generally require
Consequently, the variation of the radiative proper- an extensive and detailed library of input data, and
ties with the electromagnetic spectrum must be the calculations cannot be performed with reason-
accounted for. Spectral calculations are performed by able computational effort. On the other hand, as long
dividing the entire wavelength (or frequency) spec- as the concentration distributions of gaseous species
trum into several bands and assuming that the are not accurately known the high accuracy obtained
absorption/emission characteristics of each species for the spectral radiative gas properties from narrow
remain either uniform or change smoothly in a given band models would not necessarily increase the
functional form over these bands. As one might accuracy of radiation heat transfer predictions. Also,
expect, the accuracy of the predictions increases as it is not always convenient to use detailed, complex
the width of these bands becomes narrower. Exact models for the spectral radiative gas properties.
results, however, can be obtained only with line-by-
line calculations which require the analysis of each
discrete absorption--emission line produced as a
result of the transitions between quantized energy
3.1.2. Wide-band models
levels of gas molecules. Line-by-line calculations are
not practical for most engineering purposes but are Since gaseous radiation is not continuous but is
usually required for the study of radiative transfer in concentrated in spectral bands, it is possible to define
the atmosphere. Therefore, detailed line-by-line calcu- wide-band absorptivity and/or emissivity models.
lations will not be discussed here. The radiation absorption characteristics for each
band of any gas can be obtained from experiments
and then empirical relations can be fitted to those
3.1.1. Narrow-band models data. The profile of the band absorption may be box
Narrow-band models are constructed from spec- or triangular shaped or an exponentially decaying
tral absorption-emission lines of molecular gases by function can be used by curve fitting. These types
postulating a line shape and an arrangement of lines. of empirical models are known as wide-band models,
The shape (profile) of spectral lines is quite important and among them the exponential wide-band model
as it yields information for the effect of pressure, of Edwards and Menard 36 is commonly used. For an
temperature, optical path length, and intrinsic prop- isothermal medium, several approximate expressions
erties of radiating gas on the absorption and for the total band absorptivity and emissivity (see
emission characteristics. The Lorentz profile 32 is the Refs 37-40) as well as the reviews of the wide-band
most commonly used line shape to describe gases as models are available in the literature. 35.41 -43
moderate temperatures under the conditions of the Recently, Yu et al. 44 have devised a new "'super-
local thermodynamic equilibrium, and it is also band" model to correlate total emissivity and Planck
known as a collision-broadened line profile. 33 If the mean absorption coefficient data of infrared radi-
temperature is high and the pressure is low, the ating gases. In this model, the Edwards exponential
Doppler line profile would be more appropriate to band model has been used to approximate the
use. 33 If there are ionized gases and plasmas in the emissivities. The spectral lines of the various infrared
medium and they are influenced by interactions absorption bands of a radiating gas are rearranged
between the radiating particles and surrounding and combined into a single, combined band.
108 R. VISKANTAand M. P. M~GO~

In Figs 4 and 5, the spectral band absorptivity medium. It is clear that the relative importance of
distributions from a narrow band model a'* are short wavelength band radiation (i.e. from 1.38 pm
compared with those from a wide-band model 35 for (oJ~7000 cm - t ) and 1.89/zm (ta~5300 cm - t ) H 2 0
two isothermal media. 45 In general, the wide band bands) becomes larger as the temperature of the
model is in good agreement with the narrow band medium increases (see Fig. 5 for T = 2 0 0 0 K). The
model, especially for-2.7 /~m H 2 0 and CO2 bands error introduced by approximating the short wave-
(co=3700 cm-Z), 4.3 pm CO2 band (to~2300 cm-1), length band absorption by wide-band models is
and 6.3/~m H 2 0 band ( t a ~ 1 6 0 0 c m - ' ) ~ In these marginal, since the temperature of a typical combus-
figures, the normalized Planck blackbody function tion chamber is usually not as high as 2000 K, and
corresponding to the temperature of the medium is the other gas bands absorb radiation more strongly
also plotted to show the relative contribution of each than short wavelength bands.
gas band to the total radiation absorbed by the It should be mentioned that some of the detailed

too,o ,, '[I ~ ~ - - - r I

~.0 [i tS

Q SO.O 'l! / f v ,l .....I,


(%) : ,.,..,-

WB

2S.O "~"*"

0.0
o 2000 qooo 6o00 8000 t~>.~

w (cm-')
FIG. 4. Spectral absorptivities of H20-CO2-air mixture calculated from the narrow band (NB) and the
wide band (WB) models, spectral soot absorptivities (],,A = 1.0 x 10- 7 ma/m 3 and j~,.2= 1.0 x 10 - (' m3/m 3)
and normalized Planck's function (Iba/lha.m.): T = 1000 K, P( = 1 atm.,/M20 = Pco2= 0.1 atm., L = I m.

100.0 --

.0 /// I

a m.o I /l /
~ , ~ d~' ~ ~ ~'

H.I " Y
%, , t...: nN rl
kq: t ,,., ,:
o qooo200o ~ ~ l(XXX)
o~ ( c m " )
FIG. 5. Spectral absorptivities of H20-CO2-air mixtures as calculated from the narrow band (NB) and
wide band (WB) models, spectral soot absorptivities I]i,.t = 1.0 x 10 - 7 m3/m "*and J,,.., = 1.0 x 10- ~' m "~m "~)
and normalized Phmck's function (lh,t/lha.=**): T=2000 K, P,= 1 atm., ptt2o=Pco2=O.1atm.. L=0.5 m.
Radiation heat transfer 109

spectral properties of the combustion gases will be I

suppressed when they are combined with those of the ~= ~, as.i [ I - e - ~ , P L ] . (3.1)
i~0
particles. Because of this, use of very accurate
spectral properties of gases may not increase the The weighting factor ae,~ may be interpreted as the
accuracy of radiative transfer predictions. In Figs 4 fractional amount of black body energy in the
and 5 the soot absorptivities are plotted for two spectral regions where "gray gas absorption coef-
different soot-volume fractions. 4s Note that if ficient" xi exists, and they are functions of temper-
Iv =Jv.~ = 1.0 x 10-7, then the gas and soot absorptiv- ature. Usually the absorption coefficient for i = 0 is
ities are of the same order of magnitude, especially assigned a value of zero to account for the trans-
for longer wavelengths. However, asJ~ increases (see parent windows in the spectrum. The expressions for
the curves for J~.2=l.0 x 10-6), the soot absorption the total emissivity and absorptivity of a gas in terms
becomes dominant. The soot absorptivity also in- of the weighted sum of gray gases are useful
creases with increasing wave number, i.e. decreasing especially for the zonal method of analysis of
wavelength, since soot absorption coefficient is radiative transfer.
almost inversely proportional to the wavelength of There are several curve-fitted expressions available
radiation; we will return to this topic later. in the literature for use in computer codes. Some of
them are given in terms of polynomials 4s- 50 and the
others are expressed in terms of the weighted sum-of-
3.1.3. Total absorptivity-emissivity models gray gases. 5~-54 In only two of these expressions
soot contribution is accounted for along with the gas
A detailed modeling of the radiative properties of
contribution. 49's All of these models are restricted
combustion gases may not be warranted for the
to the total pressure of one atmosphere, except that
accuracy of total heat transfer predictions in combus-
of Leckner, 48 and all of them are for the gas
tion chambers, but definitely increase the compu-
radiation along a homogeneous path, i.e. uniform
tational effort. An in-depth review of the world
temperature and/or uniform pressure.
literature on the thermal radiation properties of
If the path is inhomogeneous then the equivalent
gaseous combustion products (H20, CO2, CO, SO2,
line model 39 or the total transmittance non-
N O and N 2 0 ) has recently been prepared,'* and
homogeneous method s5 can be used to predict
therefore the discussion will not be repeated. For
radiation transmitted along the path. However, in
engineering calculations it is always desirable to have
multidimensional geometries or if scattering par-
some reliable yet simple models for predicting the
ticles are present in the system, the use of these
radiative properties of the gases. Here, we review
models for practical calculations becomes prohibi-
some of the available models.
tive as the equations are much more complicated.
One way of obtaining radiative properties easily is
to use Hottel's charts which are presented as
functions of temperature, pressure and concentration 3.1.4. Absorption and emission coeJflcients
of a gas. '.6 Some scaling rules for the total absorp-
tivity and emissivity of combustion gases can be used The total absorptivities and emissivities are useful
to extend the range of applicability of Hottel's charts. for zero or one-dimensional radiative transfer ana-
For example, the scaling rules given by Edwards and lyses as well as zonal methods for radiative transfer.
Matavosian 47 can be employed to predict gas However, for differential models of radiative transfer
emissivity at different pressures as well as gas the absorption and emission coefficients are required
absorptivity for different wall temperatures and at rather than the total absorptivities and emissivities.
gas pressures different than one atmosphere. Of Since scattering is not important for combustion
course, in order to use these charts in computer gases (and soot particles), the gray absorption/
models, curve-fitted correlations are desirable. Other emission coefficient can be obtained from the
sources for continuous expressions are the narrow Bouguer's or Beer-Lambert's law. For a given mean
and wide band models. The spectral or band beam length Lm one can write
absorptivities from these models are first integrated ~= ( - 1/L,,,) In (1 -e). (3.2)
over the entire spectrum for a given temperature and
pressure to obtain total absorptivity and emissivity The mean absorption coefficients obtained from
curves. Afterwards, appropriate polynomials are spectral calculations as well as curve-fitted contin-
curve-fitted to these families of curves at different uous correlations were compared with measurements
temperatures and pressures using regression tech- from a smoky ceiling layer formed in a room fire and
niques. Sometimes, these curve-fitted expressions can very good agreement was found. 4a It is possible to
be so arranged that the resulting expressions would determine the so called "gray" absorption and
be presented as the sum of total emissivity or emission coefficients for each temperature, pressure,
absorptivity of clear and gray gases. These are known and path-length, which yield approximately the same
as the "weighted sum-of-gray-gases" models and are total absorptivity or emissivity of the C O 2 - H 2 0
given as '.6 mixture.
110 R. VISKANTAand M. P. MENG0~:

It is worth noting that instead of using only the decreases the gas becomes thinner, and eventually in
absorption coefficient, absorption as well as emission the limit of optically thin gas the mean absorption
coefficients should be employed. Since the total gas coefficient becomes identical to the Planck's mean
emissivity differs from total gas absorptivity, it is absorption coefficient. With an increasing size of the
quite logical to define and use two separate coeffi- enclosure, the gas becomes optically thicker and the
cients. The importance of this fact has been first mean absorption coefficient approaches Rosseland's
discussed by Viskanta# 6 He has shown that the mean absorption coefficient.
arbitrariness associated with an absorption coeffi- Planck's and Rosseland's mean coefficients are
cient can be eliminated by the introduction of a independent of the beam length and are valid only in
mean emission coefficient and a mean absorption the thin and thick gas limits, respectively. They are
coefficient, which can be related to the spectral defined as
absorption coefficient by the following definitions:
oo oo

oo oo "Kp= S K;.Ibl,dJ./ S I b~,d~ (3.5)


~s = ~ ~:a~ad)~/~ ffad2, o 0
0 0
ao

ao 1/-~a= ~(lflcz)(dlbz/dT)d,l/~(dlbz/dT)d,t. (3.6)


~,= Sgan2Ebxdg/Sn~Ebad2. (3.3) 0 0
0 o
Similar to Rosseland's mean absorption coefficient,
Here, ~a is the spectral irradiance. If the index of we can define Pianck's internal mean coefficient 35 as
refraction na of the medium is unity, then the mean 0o oo
emission coefficient will be equivalent to Planck's -~,= Sr.~,(dIb~,/dT)dg/S(dIh~,/dT)d2 (3.7)
mean absorption coefficient, s6 Also, if Ka is indepen- 0 0

dent of wavelength or the medium is in radiative


equilibrium, i.e. ffa=n]Ebx for all wavelengths, then which is also appropriate for an optically thick
the mean emission and absorption coefficients will be medium. Several other definitions of the mean
equal to each other. coefficients were discussed in greater detail by
The use of these mean coefficients is justified as Traugott. 5
long as there are no large temperature gradients in Another mean absorption coefficient was defined
the medium, s6"57 Therefore, they can be calculated by Patch s a as
separately for each zone where the temperature can oo at)

be assumed uniform. If the soot volume fraction is ~(L) = Slbagzexp(- r2,L)d2/~lb~,exp(- gaL)d2. (3.8)
high in the medium, the use of the mean absorption 0 o

coefficient would be sufficient, since ~ca (for soot


+combustion gases) would be a weak function of Unlike the first three mean absorption coefficients
wavelength. defined above, this so-called effective mean coeffi-
In order to determine the absorption and emission cient is a function of path length as it contains the
coefficients from total absorptivity-emissivity data, beam transmittance [see Eq. (2.18)] in its definition.
the corresponding mean beam length must be Therefore, Eq. (3.8) is expected to yield more accurate
properly evaluated. The definition of the mean beam predictions for the absorption coefficients of gas
length for a volume of a gas radiating to its entire mixtures having intermediate optical thicknesses
surface is given as provided that the path length is known.
It is also possible to write a mean absorption
L,,, = 4 C V/A, (3.4) coefficient based on the narrow band model of
Ludwig et al., 34
where C is the correction factor and for an arbitrary
geometry its magnitude is 0.9. 46 In general, the k~u
+~-aJ (3.9)
absorption and emission coefficients are functions of
the medium temperature, pressure and gas concen-
trations. If these coefficients are obtained from the
total emissivity and absorptivity models, they will where k, are tabulated coefficients, u is the product of
also be functions of the mean beam length. Therefore, the mean beam length and total pressure, and a is the
if the total emissivity of a gas volume is fixed, then fine structure parameter. Note that this expression is
the corresponding absorption coefficient decreases also a function of path length.
with increasing physical path length or pressure [see The mean absorption coefficients for a water
Eq. (3.2)]. The use of absorption/emission coefficients vapor-carbon dioxide-air mixture at two different
related to the mean beam length is convenient for the temperatures are presented in Fig. 6 as a function of
scaling of radiation heat transfer in practical systems, path length. 45'59 The mean absorption coefficient
As the characteristic dimension of the enclosure (~q.w,,) as calculated from Modak's model, '.9 using the
Radiation heat transfer 111

o
_ - .... -~..-~,~.. ~ ~P.~ -
.,~:.-,,i ,~ ,q
_ T=,OOOK Ke
K 2- ..~.~\ (E) Kl,n --
(~-,) "'..'.% ~,,..

I ---- '''%'%_%~ --
T= 2 0 0 0 K " . "..N,~
- .

, , I l i i i - i I t 't-v-r-~-q--=~-~
10-4 I0-s I0"2 IO-I I I0I
L (m)
F]~J. 6. Comparison of different gas aborption coefficients as a function of pathlength: P,= 1.0 alto.,
Pn,o=Pco. =0.I atm. "~

Felske-Tien 6 wide-band model, is in good agree- z0 = 3.0 m, which gives Lm= 1.08 m. The solution of
ment with the absorption coefficient calculated from the radiative transfer equation is obtained using the
the narrow-band model (Kt.,) or Patch's effective P3-approximation,61 which will be discussed in the
mean absorption coefficient (~). In this figure, xe., is next chapter. Radiative transfer calculations are
Planck's mean absorption coefficient based on the performed on the spectral basis using the wide-band
narrow-band model, 34 xe.w and ri are Planck's mean model of Edwards and Balakrishnan (see Edwards; 3s
and internal mean absorption coefficients, respec- Table X). The thirteen spectral bands used for the
tively, based on the wide-band model. 35 The mean absorption coefficient are shown in Figs 4 and 5 by
absorption coefficient xt.we is a function of path- dotted lines.
length and is calculated using Edwards' wide-band In Fig. 7 the radiation heat flux distributions
model parameters. on the cylindrical walls of the small enclosure
(L,~=0.5 m) are given for two different medium
temperatures. It is clear from these figures that the
3.1.5. Effect of absorption coefficient on the radiative
use of the Pianck mean absorption coefficient yields
heat flux predictions
about six times higher radiative fluxes compared to
In preceding sections, we have compared absorp- the detailed spectral calculations. On the other hand,
tion coefficients calculated from spectral narrow the mean absorption coefficients calculated from the
band models with those obtained from total emissiv- total emissivity model of Modak 49 yield only a small
ity models as well as with the Planck mean and overprediction of radiative fluxes in comparison to
internal mean absorption coefficients. It is also the spectral results, and the use of Planck's internal
desirable to examine the effect of different definitions mean absorption coefficients slightly underpredicts
of absorption coefficients on radiative transfer pre- the radiative flux distribution along the wall. In Fig.
dictions. For this reason, an axisymmetric cylindrical 8 the same kind of comparisons are given for the
enclosure is considered. It is assumed that the second enclosure, which has Lm= 1.08 m. Basically,
medium is a homogeneous, uniform gas ( H 2 0 - C O 2- the trends are the same as those shown in Fig. 7,
air) mixture at atmospheric pressure. The partial however, the agreement between spectral and total
pressures of water vapor and carbon dioxide are the calculations is better in this case.
same and equal to 0.1 atm., and the medium Indeed, the trends in the results predicted using
temperature is either 1000 K or 2000 K. The en- different absorption coefficients, as illustrated in
closure walls are assumed to be at a temperature of these figures, can be also deduced from the compar-
600 K and diffusely emitting, with emissivity ew=0.8. isons of the absorption coefficients given in Fig. 6.
Two different sets of dimensions for the cylindrical For example, for Lm=0.5 m, at T = 1000 K, xt.,., is
enclosure are examined. The first one has a mean somewhat larger than the x~ but it is about six times
beam length (Lm = 3.6 V/A) of 0.5 m, where ro = 0.4 m, smaller than the re. This is also evident from Figs 7
and Zo=0.9m. For the second one, ro=0.9m, and 8. From Fig. 6 we can conclude that the use of
112 R. VISKANTAand M. P. MENGOq

50 I I I I I I i

0) T= IO00K b ) T = 2000 K
250
A
40

200

x 30 / KP,w
/ \
L~. / \
\ - 150
o
"1" 20

o
I00
........ ..........
"0
o KsplctroI
n~ ~ ' ~ \Kspsctre I
lO
50

0 I I I I t , t i 0
0 I 2 0 I 2
Z/ o
FI6. 7. Comparison of radiative flux distributions on the cylidrical walls calculated spectrally and using
different mean absorption coefficients,L=0.5 m (see text for the delinitions).

50 ! I I I I I 500
a) T = I 0 0 0 K b) T = 2 0 0 0 K

A 4O 400

/
\
x 30 300

-1- 20 200
zo ,~~KItWm

spsctrel I00
spectre
S
Ki

0 I I I I I I 0
0 I 2 3 0 I 2 3 4
Z/4o
FIG. 8. Comparison of radiative flux distributions on the cylindrical walls as calculated spectrally and
using three different mean absorption coefficients,L= 1.08 m (see text for the definitions}~

Planck's mean absorption coefficient would be predictions obtained from the spectral and gray
acceptable only if the physical dimension or the total analyses. In some earlier parametric studies it has also
pressure of the system under consideration was very been shown that the change of the center and width of
small. the spectral absorption bands may yield large vari-
The spectral radiative fluxes depicted in Figs 7 and ations of the total radiative flux predictions. 57,62
8 do not always yield identical results with those Since the temperature and characteristic length of the
calculated from other mean absorption coefficients gas volume have a strong effect on both the center and
such as r,,w, or r~, and the difference between them the width of the bands, in practical systems the gas
may be as much as 100 %. Clearly it is difficult to have radiative properties are expected to show large
a simple correlation between the radiative transfer differences from location to location. Use of a single,
Radiation heat transfer 113

40 different band-width because the temperatures are


different in each zone. In the calculations an average
band-width of each spectral band was employed.
Then, the intensity of each band was adjusted
accordingly. The water vapor rotational band was
50 not included in these calculations.
In Fig. 9 curve "a" stands for the radiative heat
flux distribution obtained, including all six spectral
gas bands, i.e. 1.38, 1.87, 2.7, 6.38/~m H20 and 2.7,
4.3/~m CO2 bands. This curve is considered as the
qr 2- "benchmark" for the purpose of comparisons here.
(kW/m2) u
In order to determine whether it is necessary to
include all the bands or not, the number of spectral
bands used is reduced systematically:9 It is worth
noting that if all three minor bands (i.e. 1.38 pm,
I0 1.87 pm and 6.3 #m H20 ) are neglected the error
introduced would be on the order of 1 0 - 2 0 yo;
however, neglect of either of the major bands (i.e.
2.7/~m H20, 2.7/~m and 4.3 #m CO2) in addition to
the minor ones (see curves "c", "d" and "e") would
yield up to 50 % smaller radiation heat fluxes.
0 2 4 6 8 I0 For most practical calculations simple "mean"
z(m) absorption coefficients are widely used and preferred
F~;. 9. Comparison of radiative fluxes at the wall based on over the detailed spectral radiative properties of
spectral and mean absorption coefficient calculations. combustion gases. Therefore, it is desirable to
Water-vapor and carbon-dioxide only: "a" from all six compare the accuracy of the results predicted using
bands: "b" for 2.7 and 6.3 pm H20 and 2.7 t+m and 4.3/~m
the mean coefficients with the benchmark results. In
CO 2 bands: "c'" for 2.7/tm H20 and 2.7 pm and 4.3/~m
CO2 bands: "d'" for 2.7/~m and 6.3/~m H20 and 2.7 l~m Fig. 9, the radiative heat flux distributions calculated
CO2 bands: "'e" 6.3 pm H20 and 4.3 :~m CO2 bands; "f" for using Planck's mean absorption coefficient (curve
Planck's mean absorption coefficient;"g" Planck's internal "f"), Planck's internal mean absorption coefficient
mean absorption coefficient;+'h" for Edwards" wide-band (curve "g"), and mean absorption coefficients ob-
model.59
tained from the wide band model (curve "h") are
shown. The radiative flux denoted by curve "f" has
been multiplied by a factor of 0.4 to include it on the
mean absorption coefficient for combustion gas-
same figure; therefore, the results obtained using
mixtures, in which large temperature gradients exist,
Planck's mean absorption coefficient are not in
is not expected to predict radiative transfer realistic-
agreement with the spectral calculations. Although
ally. Consequently, gray calculations employing the
curves "g" and "h" yield 20-30 % errors in compar-
mean absorption coefficients are not recommended
ison to "benchmark" curve "a", they agree better with
for predicting radiative transfer in a medium com- the spectral results than those based on Pianck's
prised of only combustion gases, if good accuracy is mean.
required.
It is desirable to discuss the contribution of each
major CO2 and H 2 0 band on the radiation heat
3.2. Radiation Properties of Polydispersions
fluxes. Figure 9 depicts the radiation heat flux
distributions on the cylindrical wall of a combustion Analysis of radiation heat transfer in coal-fired
chamber calculated spectrally as well as using mean furnaces, combustion chambers, and other utilization
values. 59 The contributions by particles have been systems requires accounting of the effects of particu-
neglected in obtaining the results presented in this lates, such as pulverized coal, char, fly-ash and soot,
figure in order to determine the relative importance which are present in these systems. For this reason, it
of each gas band. Only water vapor and carbon is necessary to have a knowledge of the radiative
dioxide are assumed to be present. The mole fraction properties of polydispersions which, in turn, depend
distributions of these gases in the furnace were on the particle size distribution, the spectral depend-
obtained from the literature 63 for burning of low- ence of the complex index of refraction, and the
volatile coal (anthracite); therefore, the water-vapor number density for each type of particle in the
fraction in the medium was not high. The absorption combustion products. It is also necessary to know the
coefficient of the gas mixture in every zone of the spatial distribution of all the particles in the
medium is calculated from Edwards and combustion chamber. Even with all the data on hand,
Balakrishnan's wide band model. 35'39 Each spectral it is difficult and time-consuming to predict the
band corresponding to a different zone has a radiation characteristics required in radiation heat
JPgCS 1 3 : 2 - B
114 R. VISKANTAand M. P. MENGOI~

transfer analysis. Most of the time, some simplifying co ao

assumptions are made to reduce the difficulties; D,,= If(D)DdD/ If(D)dD


0 0
however, the simplifications must be reasonable for
realistic modeling of physical processes. = ZJi(Di)DiADi/~.ji(Di)&d)i (3.13a)
With the increasing coal utilization, the need for i i

radiative properties of particles formed in coal-fired


combustion systems has become more demanding. A which is also expressed as D~ o or rto if a mean radius
state-of-the-art review of the type of particles and is needed. Other definitions of the mean diameter
their effect on radiative transfer in combustion (radius) are also used in the literature, 69 including the
chambers has been given by Sarofim and Hotte164 Rosin-Rammler mean and Sauter mean, which is
and Blokh. 4 By assuming that particles are homo- given by
geneous and spherical, the radiation characteristics oo co
of a cloud of particles can be predicted from the Mie 032 = Jf(D)D3dD/Jf(D)D2dD. (3.13b)
(or Lorenz-Mie) theory. 65'66 It should be noted that 0 0

pulverized coal (char) and other particles which exist


in combustion chambers are neither homogeneous Sometimes this definition of the Sauter mean is
nor spherical. 67 Nevertheless, the extension of the modified to express it as volume to surface area ratio.
Mie theory to nonspherical (i.e. cylindrical, ellip-
soidal) particles has shown that the radiation
characteristics of a cloud of irregular shaped particles 3.2.1. Types and shapes of polydispersions
are not very sensitive to the geometrical shape of the
particles. 66,6a Therefore, the use of the equivalent In combustion chambers, soot, pulverized coal,
spherical particles assumption and the Mie theory char, and fly-ash are the polydispersions to be
for coal combustion systems appears to be a considered. Soot is one of the most important
reasonable compromise. In this section the method- contributors to radiation heat transfer in practical
ology for calculating the radiative properties of systems. Typical diameter of the soot particles is
polydispersions is given. Some simple expressions about 30 nm to 65 nm, 64 yet the sizes of soot
are also suggested for use in practical calculations. aggiomorates may be much larger.'* Mainly because
Following the Mie theory, the spectral absorption, of the small size of the soot particles, scattering of
extinction, and scattering coefficients needed for radiation by soot is negligible in comparison to
radiative transfer analysis can be evaluated from the absorption, and its radiative properties can be
equation, calculated easily provided that the complex index of
refraction and volume fraction distribution data are
oo available. Numerous experimental studies (see, for
qa(ha,N) = I Q~(D,A,haX1tD2/4)f(D)NdD, (3.10) example, Refs 4, 70-73 for citations) have reported a
0
complex index of refraction as well as volume
fraction data of soot. Recently, Felske et al. 74
where qa stands either for spectral extinction coeffi-
discussed the effect of different soot particle shapes
cient fla, the spectral absorption coefficient xa, or for
on the scattering characteristics of radiation and
the spectral scattering coefficient tra, and Q, is the
presented a framework for determining the character-
corresponding efficiency factor which is a function of
istics of soot agglomerates using those of spherical
the size (diffraction) parameter (x=nD/2) and the
particles. They demonstrated the sensitivity of the
wave-length of radiation 2. Here, h a is the refractive
soot radiative properties on the inhomogeneity of
index of particles, N represents the particle density,
the particles by using the coated sphere model (see
and f(D) is the normalized size distribution function,
Subsection 3.2.3).
oo The spectral complex index of refraction is the
f(D)dD= 1. (3.11) most fundamental optical property required to
0 calculate radiative characteristics of polydispersions.
It is computationally time-consuming to take into
For most practical problems, a discrete size distri- account the dependence of the index of refraction on
bution of polydispersions is required. Hence, it is wavelength, but conceptually it is straight-forward.
better to replace the integral of Eq. (3.10) by a finite In the literature, there are some published data
series. Then, using the "step-size distribution", the for the complex index of refraction of various
radiative properties can be expressed as coals. 69,75,76 An extensive compilation of the com-
plex index of refraction data, including that for
different Soviet Union coals, is also available.4 The
qa(na,N) = ~Q,v,, (D,,2,haX~D~/4)fi(D,)NAD,, (3.12)
1
early experiments for determining ha were usually
based on the "Fresnel reflection" method. More
where i designates the diameter ranges. The mean recently, Brewster and Kunitomo 76 proposed a new
diameter of particles in the cloud can be obtained as method, the so-called "particle extinction" technique.
Radiation heat transfer 115

Their results show that there is approximately one real and the imaginary parts of the refractive index
order of magnitude difference between the impinging on composition were also examined. A semi-empirical
part of the complex index of refraction measured mixture rule was developed to allow prediction of the
with these two techniques. In brief, there are large real part of the refractive index from 1 #m to 8 #m in
differences between the reported spectral data for terms of the weight percents of the major oxide
the complex index of refraction of coal particles components SiO2, Al2Oa, CaO, MgO, TiO2, and
reported by different investigators, 4'69'7'~-76 and, Fe203. The mixture rule is based on the refractive
therefore, more research attention is needed in this indices of the pure oxide components, with two small
area. It is also bdieved 64 that the radiative proper- modifications to improve the agreement with the
ties of char particles do not show distinctive measured refractive index data.
differences from those of other pulverized-coal Shape of a particle is another important indepen-
particles. Unfortunately, to the authors' knowledge, dent parameter that should be considered in pre-
there is no fundamental study which supports this dicting the radiative properties. For the particles in
conclusion for various coals and at different wave- combustion chambers, it is difficult to imagine a
lengths of radiation. single, unique shape. Usually shapes of pulverized-
The contribution of fly-ash particles to radiation coal particles or soot agglomerates are irregular and
heat transfer in pulverized-coal flames exceeds that of random; yet, sometimes, surprisingly uniform and
combustion gases or soot substantially; 4 therefore, simple shapes are observed. For example, fly-ash
special attention must be given to the radiative particles from coal-fired boilers show fairly smooth,
properties of these particles. Although limited, some spherical shapes, a4"8~ The soot, on the other hand,
data for radiative properties of fly-ash particles have may agglomerate to form relatively long tails of radii
been reported in the literature. 4"77- s3 on the order of the coal particle radius due to the slip
The refractive index of fly-ash is sensitive to its velocity between the coal particle and surrounding
chemical composition, and this is attributed primarily gases. 86's7 These tails can be considered as infinitely
to the varying amounts of oxides of silicon, alumin- long cylinders. The simple shapes are most desirable
ium, iron and calcium (i.e. SiO2, A1203, Fe203 and for the simplicity of calculations as the computational
CaO) in the ash. The experimental studies have effort is reduced significantly for uniform, symmetric
shown that the index of refraction of different fly-ash shapes. However, a large fraction of particles sus-
samples from the same flame may be drastically pended in combustion products have totally irregular
different, probably indicative of the microscopic shapes. Experimental measurements show that there
conditions for their formation. 77 According to Wall are some differences in the scattering properties of
et al. 78 the complex refractive index of fly-ash is in these particles in comparison to Mie theory calcu-
the range from ha= 1.43 -0.307i to ha= 1.50-0.005i. lations, as where for irregular shape particles: (a)
These numerical values of the imaginary part of the oscillations of efficiency factors vs angle and vs size
complex index of refraction correspond approxi- parameter are damped; (b) more side scattering
mately to the values measured by Blokh, 4 whereas (60-120 ) is observed; (c) less backscattering is
the values of the real part of the complex index of observed and; (d) the agreement with Mie theory
:refraction are somewhat lower than those reported. becomes worse for other radiative properties as the
The imaginary part of the refractive index of fly-ash size parameter increases past x = 3 or 5. sa For a cloud
particles formed during combustion of pulverized- of irregular shape particles, however, the observed
coal in a fluidized-bed furnace was of the order of differences in comparison to those for spherical
0.01. 76 This clearly indicates the uncertainty in the particles are less significant. 66'a8
complex index of refraction of fly-ash particles
formed in pulverized-coal combustion systems.
3.2.2. Prediction methods of the particle radiative
Recently, Goodwin 83 has reported extensive results
properties
of an experimental study of the bulk optical con-
stants of coal slags. The effects of chemical compos- When radiative properties of particles are needed,
ition, wavelength, and temperature were examined. the following quantities, arranged in order of
Both synthetic slags, prepared from oxide power increasing complexity are to be considered: 8s (i)
mixtures, and "natural" slags, prepared by re-mdting extinction cross-section, (ii) scattering cross-section,
fly-ash or gasifier slag, were used. Transmittance and (iii) absorption cross-section, (iv) single-scattering
near-normal reflectance measurements were made on albedo, (v) radiation pressure cross-section, (vi)
their polished wafers cut from the slags, from which asymmetry factor, (vii) unpolarized phase function,
optical constants were determined. The imaginary (viii) Legendre coefficients of unpolarized phase
part of the refractive index was shown to depend function, (ix) parallel and perpendicularly polarized
primarily on iron, silica and OH content of the slag. scattered intensities, (x) Stokes parameters, (xi)
Iron is primarily responsible for absorption in the Mudler matrix, and (xii) Legendre coefficients of
short-wavdength infrared region (1 #m<3.<4/Jm), Mueller matrix dements. The last four quantities in
and silica is responsible for absorption at longer this list may not be critical for studying radiative
(.k> 4/~m) wavelengths. The dependences of both the transfer in combustion systems. However, the other
116 R. VISKANTAand M. P. MENG0~

quantities are definitely needed for radiation heat with the complexity (or asymmetry) of the shape of
transfer calculations. the particles. Recently, Wiscombe and Mugnai as.l i l
One of the most extensively used models to predict developed a vector algorithm for the EBCM code of
the radiative properties of particles is the Mie Barber t4 and obtained the scattering properties for
theory. 65'~6 Although it is widely known by this various axisymmetric particles whose shapes are
name after Mie's exact solution of Maxweil's equa- determined from Chebyshev polynomials. Their
tions for the scattering of an incident plane wave on results show that there are significant differences
a sphere, s9 the solution was also obtained independ- between the radiative properties of spheres and
ently by Lorenz and Debye (see Kerker 66 for detailed arbitrary shaped particles depending on the irreg-
historical discussion). The exact solution for a right ularity of the surface characteristics. The compu-
circular cylinder with radiation incident normal to tational time required for these calculations is too
the cylinder axis was given by Rayleigh. Basically, in formidable as to justify the extensive use of the T-
the Mie theory the vector Heimholtz equation is matrix method for practical problems.
solved exactly by expanding the electric field in an
infinite series of eigenfunctions. In general, these
series are double series, and they are not easy to
3.2.3. Simplified approaches
evaluate; however, for spheres and infinitely-long
circular cylinders they can be reduced to single series, One of the simplifications usually made in calcu-
and exact solutions can be obtained. The Mie lating the radiative properties of particles is related
theory for spheres has been treated extensively in to their shape. If it is possible to assume that the
the literature, 65'66'9 and some formulations for particles are spherical, then exact solutions from Mie
cylinders,9-9'* for elliptic cylinders 95 and for theory can be obtained effectively and with much less
spheroids 96 have been given. There is no need to computational effort in comparison, for example, to
repeat the details of Mie theory here; the interested the T-matrix method. The properties of irregular
reader is referred to one of the classical refer- shaped particles can be obtained by assuming them
e n c e s 6 5 ' 6 6 ' 9 0 o n the subject. as equal-volume spheres if the size parameter
The Mie theory has been used extensively, es- (x = riD~A) is small or equal-projected-area spheres if
pecially during the last two decades, with the help of the size parameter is large, as The nonsphericity of
the computer algorithms which have been devel- particles can be traded off against inhomogeneity by
oped 9 0 ' 9 7 - 9 9 a s well as widespread use of digital assuming that the index of refraction varies from the
computers. Its restriction to simple, smooth particles core to the periphery. 66 By picking a functional form
has led researchers to investigate some other possi- for this variation that allows a reasonably simple
bilities to model the scattering of radiation by radial solution with one or two adjustable para-
irregular shaped particles. Several new approaches to meters, it may be possible to match nonspherical
the solution of the problem have been proposed over scattering properties. Then, the solution for an
the years, including exact differential equation inhomogeneous sphere can be obtained rather than
approaches ~oo.lot as well as exact integral equation for an irregular-shaped particle, and this is signifi-
methods. 12-~* In addition to these, there are cantly simpler. It is also worth noting that the effect
several approximate techniques available, including of shape becomes less critical if there is a size
the geometrical theory of diffraction ~s for pre- distribution of particles, as size-averaging in ob-
dicting the scattering by sharp-edged particles; the taining the radiative properties "washes out" the fine
method of moment for scattering by a perfectly details of nonspherical scattering. 66.as
conducting body; 16 as well as perturbation17 and The Mie calculations for the efficiency factors of
point matching methods ls for nearly spherical spheres are relatively less time-consuming and easier
particles. Some empirical models have also been to use than the other exact models. However, the size
proposed and shown to be very accurate provided of the particles in combustion chambers are func-
that some experimental data are available, t9 The tions of time and space, and the properties must be
details of these methods and others can be found in calculated for each new set of size distributions. In
the literature, ss'9A t o multidimensional and spectral radiative transfer
Among these models, the integral equation method analyses use of Mie codes for this purpose is
or as more widely known, T-matrix or extended impractical. Because of this, it is desirable to have
boundary condition method (EBCM), 1oz- ~o, seems simple approximations for the efficiency factors. One
to be the most promising as it is capable of solving such approximation has been given by Mengii~ and
the scattering of radiation by any irregular shape Viskanta, lt2 where the efficiency factors for poly-
particle. In the EBCM, the incident and scattered dispersions are obtained starting from the anomal-
electric fields are expanded in vector spherical ous diffraction theory ~5 and are expressed in conven-
harmonics, and then by making use of analytic ient, closed form. In Fig. 10 the Mie theory
continuation techniques the integral representation predictions for the normalized extinction and scat-
of the fields is reduced to a set of linear algebraic tering coefficients are compared with those of the
equations. The complexity of these equations increase simplified model, and in Fig. 11 the predictions of
Radiation heat transfer 117

x F(~)

I0 ~ tO-t I0 0 tO i I0 e IO : I0 q tO 6
10-7 ........ I ........ i ' ' ' ..... I ........ I ' ' ..... I ........ I
...... 10-7

10-.8 lO-a
CP,RBON
a P,NTHRRCI TE
[n~q + BITUMINOU$ [m"]
10-e x BITUMINOUB-K i0-9
LIGNITE
FLY-fiSH
lo-iO i0-I0

lO-ii ........ i ........ i ........ i ........ I ........ , ........ i ...... lO-ii

lO-t lO o 10 I 10 ! lO 3 10 ,I lO 5 10 e

xF(~)

FIG. 10. Comparison of Mie theory results (points) for the normalized extinction and the scattering
coefficients with those calculated from tin approximate analysis (lines).tj 2

1.000
c) CRRBON
0.800
A P,NTHRRCITE
+ BITUMINOUS
COx
X BITUMINOUS-I(
0.600
,i.~ ,..OOO& LIGNITE
+ FLY-fiSH

........ I
I0 "2 I 0 -i IO o 10 i I0 ~ I0 : IO N

x F('~)

FIG. 1I. Comparison of Mie theory results (points) for the single scattering albedo with those calculated
using approximate analysis (lines).~t 2

the scattering albedo from the Mie theory and the in the volume fraction of polydispersions and the
simple model are given.l~ 2 In these figures, flz and aa complex index of refraction data, the agreement
are normalized spectral extinction and scattering between the model and exact calculations appears to
coefficients, respectively. The normalization factor is be remarkably good, and, because of this, these
NxF(ha), with N being the number of particles per simplified models would be useful for radiation heat
unit volume, x is the size parameter, and F(h,t) is a transfer calculations in combustion chambers. Note
function of the complex index of refraction. Note that the single scattering albedo 09 is related to
that absorption, extinction and scattering coefficients by

F 24naka ] <.o = o/,a = 1 -- ~://7. (3.15)


Q,=xF(ha)=x 2 ~ 2 2 (3.14)
L ( n a - k a + 2 ) +4naka.J
In the literature, there are also some empirical
is the absorption efficiency factor for very small size relations available for the radiative properties of
spherical particles (x--g)) as obtained from the polydispersions. Buckius and Hwang 113 calculated
Rayleigh limit of the Mie theory. The discrete points absorption and extinction coefficients as well as the
shown in Figs 10 and 11 are the results obtained asymmetry factor of several coal polydispersions
from the rigorous Mie theory for the corresponding using Mie theory and showed that they were almost
index of refraction of specific particles. The lines are independent of the size distribution and were
from the analytical, closed form expressions given by functions of average radii r32 [see Eq. (3.13b)] and
Mengii~ and Viskanta. 112 Considering the uncertainty the complex index of refraction. They obtained some
118 R. V[SKANTAand M. P. MENG0(;

10-2 , , ,

. . . . '~ -'~. . . . J. . . . . . . . . . . . . . K.'lX,) "IN,, ~ .

i0"s I0"2

10"4 ~ i0"3

i0-4
I0x I02 I03 i04
r,zT [/.t.m K]
FIG. 12. Phmck and Rossehmd mean coefficients for coal. The shaded area represents results for w~riations
in temperature between 750 and 250 K and three coals. ~ 3

empirical correlations for the radiative properties of worth noting that the anomalous diffraction theory
coal particles which could be readily used for used for spheres also yielded accurate and simple
predicting radiation heat transfer in coal-fired com- relations (see Ref. 112). Most recently, Mackowski et
bustion systems. Also, they plotted the normalized a/. 117 derived the same kind of relations for the
Planck and Rosseland mean absorption and extinc- spectral radiative properties of cylindrical soot
tion coefficients as functions of the mean radius- agglomerates. They showed that small size cylin-
temperature product (Fig. 12) and obtained some drical particles extincted radiation two to five times
empirical relations for these coefficients. As seen more than spheres. At large radii, on the other hand,
from this figure, for small radii particles, extinction the ratio of cylindrical extinction and absorption
and absorption coefficients are identical; however, coefficients to those for spherical particles approach
with increasing radius the scattering of radiation constant values regardless of the wavelength of
also becomes important, and fl and x diverge from radiation} 17 Also, some empirical relations similar
each other. Viskanta et al. 1~4 aIso obtained similar to those obtained for spherical particles are pre-
results and discussed the effects of several indepen- sented. It is also possible to extend the relations to
dent parameters, such as size distribution, coal type mixtures of different types and shapes of particles
and wavelength of radiation on the radiative proper- using the T-matrix method. For a specific (coal)
ties of polydispersions. It is worth noting that combustion problem, a library of empirical relations
although different definitions of mean radius are can be constructed. The use of these relations will
used in these studies, i.e. rio [see Eq. (3.13a)] 112 and speed up the calculations significantly, since there
r32 [see Eq. (3.13b)], 11a'1~4 still similar results will be no need for lengthy and time consuming Mie
independent of size distribution are obtained. This or T-matrix method calculations.
indicates that a polydispersion can be often des- When the size parameter ( x = n D / 2 ) becomes
cribed by a weighted particle radiusJ 15 vanishingly small (x-*0) the size of the particle
All of the studies discussed above used the becomes less important. In this limiting case, the
spherical particle assumption in obtaining the re- absorption efficiency factor is a function ofx [see Eq.
lations for radiative properties of particles. Perfect (3.14)], whereas the scattering efficiency factor varies
spheres are not encountered in nature, and, therefore, with x 4, such as
it is desirable to obtain similar relations for other
than spherical shape particles. Stephens ]~6 has r~-I 4
shown that the anomalous diffraction theory devel- 0s = 3 ~ X . (3.16a)
oped by van de Hulst 65 can be extended to infinite-
length cylinders. The absorption and extinction
efficiency factors calculated from this simplified The extinction efficiency factor is written as
theory are in good agreement with those obtained
from a rigorous solution of Maxwell's equations. It is Q,. = Q,, + Qs. (3.16b)
Radiation heat transfer 119

These expressions are obtained from the Rayleigh 3.2.4. Scattering phase function
limit of the Mie theory. 6s Here, ha=na-ika is the
complex index of refraction. It is worth noting that In modeling radiation heat transfer in a partici-
with decreasing x (or D), the scattering efficiency pating medium, the scattering of radiation by
factor becomes negligible in comparison to the particles must be properly accounted for. This
absorption efficiency factor. Indeed, these expres- requires the use of the scattering phase function
sions yield the extensively used soot absorption (scattering diagram), which represents the probability
coefficient, such as that radiation propagating in a given direction is
scattered into another direction because of the
inhomogeneities and/or particles along the path of
xa = 7f J2 (3.17) radiation. In combustion chambers, the scattering of
radiation takes place mainly because of the particles.
wherefv is the volume fraction of soot particles and The phase function, along with other radiative
the value of "7" was suggested by Hottel and properties, such as absorption, extinction and scat-
Sarofim 46 for typical soot particles observed in tering coefficients, can be obtained either exactly
combustion chambers. After studying the available from the solution of Maxweli's equations for spher-
experimental data for several flames Siegel l~s has ical or infinite-length cylindrical particles 65,66'9 or
shown that the coefficient in Eq. (3.17) is between 3.7 from some approximations such as the extended
and 7.5 for coal flames; 6.3 for oil flames, and 4.9 and boundary element method (EBCM) for arbitrary
4.0 for propane and acetylene soot, respectively. A shaped particles ~2- lo4 as functions of wavelength,
detailed discussion of the spectral and total absorp- characteristic particle dimension and complex index
tion characteristics of uniform-diameter, spherical of refraction. The phase function is written as
soot particles covering a very wide range of sizes
(0.001 <D<10/~m) is given by BlokE'* Oa(g'---,g) = 4la(-g)/x2Qs (3.19)
Equations (3.14) and (3.16) were obtained for
spherical particles; therefore, the approximation
where la(g) is the incident radiation intensity and x
given by Eq. (3.17) may not yield accurate results
is the size parameter with the effective diameter D
for arbitrary shaped small particles. 1~ For non-
and radiation of wavelength 2. Note that the
spherical particles, an expression for the average
radiative properties are not only functions of the size
absorption efficiency factor was derived by inte-
grating over a distribution of shape parameters in of the particle or the wavelength of radiation, but
the Rayleigh-ellipsoid approximation, 119 such as they are functions of the size parameter x, which can
be considered as a scaling factor. In Eq. (3.19) Qs is
the scattering efficiency factor, which is defined as 9
2ha
Q,=xF(ha)=x Im [hA-- 1 (log na-ika)] (3.18) Qs = CJA = (Ws/I,)/A (3.20)

where x depends on the effective diameter D(--- V/A). where A is the particle cross-sectional area projected
This relation yielded very good agreement with the onto a plane perpendicular to the incident beam li
experimental data for quartz particles.119 (e.g. A = nD2/4 for a sphere of diameter D); Ws is the
In Table 2 a comparison is given of spectral F(ha) energy scattering rate by the particle, and C s is the
functions for spheres [see Eq. (3.14)1, infinite-length scattering cross-section. Similar expressions can be
cylinders 117 and ellipsoids [see (Eq. 3.18),1 at four written for extinction and absorption efficiency
different wavelengths. It is important to note that the factors by replacing the subscripts "s" in Eq. (3.20) by
results for ellipsoids are between those for spheres "e" for extinction and "a" for absorption. These
and cylinders. The spectral complex indices of quantities are obtained from the Mie theory or
refraction used in this comparison are from the approximate models. 9
dispersion relations developed by Lee and Tien 71 for Use of the phase function in the form of Eq. (3.19)
acetylene and propane soot at 1700 K. would be a very time consuming procedure, A more
convenient form of the phase function is obtained by
expanding it in a series of Legendre polynomials) 20
TABL[2. Comparisons of spectral F(ha)functions for N
different shape small soot particles a(W) = ~ a,.~P,(~P) (3.21a)
rl=O
/(/am) na ka F,~,,, F~,n~r Fomt,,ola
0.50 1.92 0.55 0.754 1.871 1.700 where
1.50 1.88 0.73 1.007 2.450 2.134
2.50 2.10 1.09 1.140 3.683 2.888 1
5.00 2.69 1.57 0.863 6.073 3.888 an,a-2n+l J" ~a(W)P,(~)dD (3.21b)
O=4f
120 R. VISKANTAand M. P. MENO0q

is the expansion coefficient and P, is the Legendre 1


polynomial of degree n. By changing the upper limit A=U~ I o~(,I,)dn
t~=2~
N of the series, any phase function can be written in
the form of Eq. (3.21). The coefficients a,.a can be =+ ~ (-1)"a2"+'(2m)!
obtained by employing the orthogonality relations of (3.26a)
r.=0 22"+im!(m+ 1)t
Legendre polynomials. In order to accurately repre-
sent the phase function of highly forward scattering
ba = 1 - f a . (3.26b)
particles, however, as many as 100 terms may be
required in the series. For the multidimensional
The factors ba and fa are especialy useful when
radiative transfer calculations, the use of such a obtaining solutions of the radiative transfer equation
complicated scattering phase function is not prac- using flux methods.
tical either. Consequently, some further simplifi- Brewster and Tien 125 have given a different
cations are required. If N = 0 , the phase function is definition of the backward scattering coefficient for
written as
an azimuthally symmetric layer such as
(IDa(W)= 1 (3.22) 1 o
Ba=~ ~ ~a(/~,/~')d/~'du (3.27)
which is for isotropic scattering. If N = 1, then the 0 -1

linearly anisotropic scattering phase function is


obtained, This expression is valid for a plane-parallel layer of
particles, whereas Eqs (3.26) are appropriate for
~ a ( ~ ) = 1 + a l,acos W. (3.23) scattering from a single particle. It has been shown
that for a cloud of non-absorbing spherical particles
For N = 2 the phase function corresponds to second- (with h=1.33, x=6.0), b~(=0.036) is drastically
degree anisotropic scattering, and if the expansion smaller than Ba(= 0.137).
coefficients are set arbitrarily such that at.,t=0 and In atmospheric studies, the Henyey~3reenstein
a2,a= 1/2, this yields the Rayleigh scattering phase phase function approximation is often used 121 and is
function.~ 9 expressed as
Most of the particles encountered in combustion
chambers (pulverized coal, char or fly-ash) scatter
radiation predominantly in the forward direction. @n-o,a( W)= [1 +O]-2gacos tF]3/2" (3.28)
Such a scattering behavior can be modeled using a
Dirac-delta function. The transport, delta-M and Here, g~ is the asymmetry factor and is defined as
the delta-Eddington approximations are of that
form. T M The delta-Eddington approximation is
written as ~22 ga= ( c o s W ) = S la(g)cosWd~/ S la(g)df~ (3.29)
fl=4x O=4f

~a(W) = 2f~6(1 - cos W) + (1 -faX1 + 39acos W) (3.24) which can be directly obtained from the Mie theory.
Although it approximates the Mie phase function
where fz and 0a are related to the expansion quite accurately, the application of the Henyey-
coefficients defined by Eq. (3.21b) as 59 Greenstein phase function approximation to multi-
dimensional geometries may be quite tedious.
fa = {i:al i f_ ~ 2)12 i i : : ,~ 1)/2 (3.25a) Several different approximations for the scattering
phase function, such as linearly anisotropic scattering,
delta-M, delta-Eddington, transport or Henyey-
and Greenstein approximations, have been reviewed in
detail by McKellar and Box)21 They have concluded
a l ,,a - f a that for highly forward scattering particles the
9a = (3.25b) delta-Eddington approximation ~22 is the most
1-fa
accurate and the simplest of all the approximations
provided that al,a>a2,a. A detailed account of mentioned. In modeling radiative transfer in coal-
Dirac-deita phase approximations has recently been fired furnaces the delta-Eddington approximation
given by Crosbie and Davidson? 23 for the scattering phase function is desirable for two
In the heat transfer literature, another phase reasons: (1) it represents the highly forward-directed
function approximation has found wide application. scattering of radiation by the pulverized coal and fly-
The phase function is expressed in terms of the ash particles; and (2) it is compatible with differential
forward (fa) and backward (ba) scattering coefficients, approximations such as the spherical harmonics
and they are written in terms of a.'s of Eq. (3.21b) for approximation used to model the radiative transfer
an azimuthally symmetric medium such as 124 equation.
Radiation heat transfer 121

The scattering phase function of particles is the concept may be of limited utility for predicting
directly related to the size (or diameter) of the radiation heat transfer in multidimensional combus-
particles. Therefore, for polydispersions there should tion systems which contain particles, If the emissivity
be as many scattering phase functions as the number (or absorptivity) of a particle laden flame is known
of size intervals considered. For the sake of simplicity, then the extinction coefficient of the medium can be
it is desirable to have a single, mean scattering phase written as
function over the entire particle size range. Then, the
mean scattering phase function can be written as
fla.tot= - L~ In (3.36)
1 N
~a = : - ~ a~,i Pa.~, (3.30)
0"2 i provided that the mean-beam-length L,, is known.
However, fla,tot and toa are interrelated properties
where N is the number of the intervals. If the [see Eq. (3.34)]. If the mean coefficients are to be
delta-Eddington phase function approximation is used, the equations given above should be rewritten
used for the scattering phase function, the corres- by dropping the subscript 2, and appending the
ponding mean parameters are defined similarly, appropriate mean coefficient subscript.
These definitions require the mean beam length of
1 N 1 N
radiation Lm, which is a vague concept. 59 It is
.'/-~- ~- ~O'2.,if]i, "g~-~- ~- EU2..iOa.i. (3.31)
G~ i G'l i
defined 2 as a radius of a gas hemisphere which
radiates a flux to the center of its base equal to the
average flux radiated to the area of interest by the
actual volume of gas. Although the concept yields
3.3. Total Properties accurate results for simple systems, for complicated
geometries it needs additional research attention.
Once the absorption, scattering and extinction
Recently, Scholand and Schenkel ~27 have calculated
coefficients of polydispersions, such as pulverized
coal, char, fly-ash, and those of soot and combustion the mean beam length of radiation between a volume
element and the surfaces of rectangular paral-
gases are known, the total radiative properties are
lelepiped enclosures. Cartigny ~28 has extended the
written as
definition of the mean beam length to an optically
thin scattering medium, which can be used for
ra.tot = ~xa.~ly-i + xa.,,**,+ ~xa.,,,,,_.i, (3.32) calculation of radiative transfer in sooty flames.
i j The empirical relations for the total mean extinc-
tion and absorption coefficients for fly-ash, pulver-
ized coal and char particle polydispersions have also
fla,to,= xa.tot+ ~aa.vo,y-i, (3.33)
i been reported. 4 It has been found, for example, that
the mean extinction coefficient fl of fly-ash can be
and expressed by an empirical equation of the form,

fl= g,, F(CL)ApC (3.37)


o93= ~tra,~ly_ i/fl~.,tol (3.34)
i
where Q,, is the extinction efficiency factor; F(CL) is
where "poly-i" refers to i-th type polydispersion, and the function which accounts for the dependence of
"gas-j" refers to j-th gas species. Note that if there the extinction coefficient on the product of the
are no scattering particles in the medium, then concentration C and the layer thickness L, and At, is
fla.tot= K~.,tot and t~a = 0. the surface area of a particle. The total extinction
An alternative formulation of total absorptivity efficiency factor Q~ has been found to depend, as
and emissivity of a scattering medium has been might be expected, on the type of coal burned, fly-ash
recently given as ~26 particle size and the spectral distribution of incident
radiation determined by the black body temperature
F tg'l''tt [1 --exp(-fl~.,totLm) ] . (3.35) T used as the radiation source. Based on experi-
~. = C~= Lfl,l.,tot -I mental data it is possible to express Q,. by an
empirical equation,
In writing this expression it was assumed that
spectral irradiance was equal to the spectral emitted Q,.=0.07 A(xT) 1/3. (3.38)
flux of the surroundings. The spectral absorptivities
of polydispersions of coal and fly-ash particles have The empirical constant A depends on the type of coal
been predicted using a two-flux approximationJ ~4 burned and the shape of the fly-ash particle, and x is
The expression for the absorptivity 1~4 is not as the size parameter based on the mean particle
simple as that given by Eq. (3.35). This suggests that diameter. The function F(CL) has been determined
122 R. VISKANTAand M. P. MENG0~'

empirically and was found to depend on the type of 4.1. Exact Models
coal burned. 4
The total effective absorptivity of a fly-ash layer of The most desirable solution of any equation is its
thickness L calculated from the expression exact closed form solution. The exact solution of the
integrodifferential radiative transfer equation can
cZfe= 1 - exp( - ~L) (3.39) only be obtained after some simplifying assumptions,
such as uniform radiative properties of the medium
has been found to agree well with the experimental and homogeneous boundary conditions. For one-
data.'* It was determined from the data that the dimensional, plane-parallel media, exact solution of
optical thickness zL(= ~L) of the layer varies linearly the RTE has received much attention in the atmos-
with CL only for moderate values of CL(<20 g/m2). pheric sciences, 12a 5,t 6 neutron transport 13~- ~33 and
At higher values of CL the mean extinction coef- heat transfer ~9'2'~34 literature. A detailed review of
ficient ~ starts to depend on CL, because the one-dimensional exact solution methods is avail-
radiative properties of fly-ash particles depend on able? 35 However, there have only been a few
wavelength. This leads to the departure of the attempts to formulate and solve the RTE for
function zL(CL) from linearity. multidimensional geometries.
The Hottel charts for the emissivity and absorptiv- One of the earliest accounts to formulate the
ity of combustion gases are very convenient for radiative transfer equation in a three-dimensional
practical calculations. Skocypec and Buckius ~29 space with anisotropic scattering was that of
and Skocypec et a/. 13 extended these charts to Hunt? 36 He considered a phase function comprised
include isotropically scattering particles. In their of three terms in Legendre polynomials and reduced
calculations, they obtained the gas properties from the integrodifferential radiative transfer equation to
the Edwards wide-band model 35 and presented an integral equation. Cheng ~37 used a rigorous
hemispherical emissivities in graphical form and approach to solve the RTE for an absorbing-
discussed the effects of optical thickness, pressure, emitting medium in rectangular enclosures, and Dua
temperature and single scattering albedo. These and Cheng ~38 extended this method to cylindrical
charts yield accurate radiative properties without geometries. For an absorbing, emitting, and scat-
any additional calculations; however, they cannot be tering medium Crosbie and his co-workers presented
used directly for predicting the local radiation heat exact formulations of the RTE for three-dimensional
flux in a combustion system. rectangular ~39 as well as three-dimensional cylin-
dricaia4 enclosures. The solution of these equations
for cylindrical geometry was obtained by the method
4. SOLUTION M E T H O D S of subtracting the singularity? 4a The exact solutions
of RTE for an absorbing and emitting medium were
The radiative transfer equation is an integro-
also solved by Selcuk ~42 in a three-dimensional
differential equation, and its solution even for a one-
rectangular enclosure employing a numerical scheme.
dimensional, planar, gray medium is quite difficult.
In a cylindrical geometry, the radiative transfer
Most engineering systems, on the other hand, are
equation is obtained from Eq. (2.11). Then, the
multidimensional. In addition, spectral variation of
integral form of the source function, for an absorbing,
the radiative properties must be accounted for in the
emitting and isotropically scattering medium with
solution of the RTE for accurate prediction of
incident diffuse radiation source on one of the end
radiation heat transfer. These considerations make
surfaces of the cylinder can be written as 14
the problem even more complicated. Therefore, it is
almost necessary to introduce some simplifying
S2(r,z,q~)= (1 -- 0)a)/ba[ T(r,z,q~)]
assumptions for each application before attempting
to solve the RTE in its general form. It is not possible
0) 2 r 2x
to develop a single general solution method for the + - - J S l d,~(r',d?')e-P~"z'~)X; (x~ ) - 3zr'dq~'dr'
equation which would be equally applicable to 4no o
different systems. Consequently, several different
solution methods have been developed over the
+m2 z~ r~ ~Sa(r,,z,,(a,)flae_Pa%x~ 2r,dda,dr,dz ' (4.1)
years. According to the nature of the physical system, 4nooo
characteristics of the medium, the degree of accuracy
required, and the available computer facilities, one of where
several different methods can be adopted for the
solution of the problem considered. Before choosing x ; = [r 2 + (r') 2 - 2rr'cos(q~ -- qS')+ z2] 1/2 (4.2a)
one solution method over another one, it is import-
ant to know the advantages and disadvantages of Xp = [r 2 + (r')2 - 2rr'cos(cb - ~b')+ (z- z')2] I/2. (4.2b)
each method. In this section, several radiative
transfer models of interest to combustion systems are Here, the primes are used to denote the dummy
discussed, and their features are highlighted. variables, and l~,a is the spectral diffuse radiation
Radiation heat transfer 123

source on the end of the cylinder at z=O. Note that Monte Carlo method can be used for any complex
ld. a can also be interpreted as the diffuse emission geometry, and spectral effects can be accounted for
and reflection from the walls. Some additional without much difficulty. Mainly for this reason, the
integral terms are to be added to these expressions to method has been used extensively in atmos-
account for the other surface effects` After some pheric 143'144 and neutron transport 133 studies. It
lengthy and tedious algebra, the implicit expressions has also been successfully employed to solve some
for the radiative fluxes in the r, z, and ~b directions general radiation heat transfer problems 14~,z46 as
can be derived: 14 well as radiative transfer problems in multi-
dimensional enclosures ~,17 and furnaces. ~4s,1,,9

F,.a(r,z,ck) = ,[ 2[ld,,t(~,~b')-P'~;(x;) -~
There is no single Monte Carlo method. Rather,
there are many different statistical approaches. In its
00
simplest form, the method consists of simulating a
z r 2z finite number of photon (energy packet) histories
x J r - r'cos(~b - ~b')]zr'dc~'dr'+ [ [ I Sa(r',z',c~') through the use of a random number generator. 133
0 0 0
For each photon, random numbers are generated and
used to sample appropriate probability distributions
x flae-P~%(xp)- air - r'cos(~b - t~')] r'dq~'dr'dz' (4.3)
for scattering angles and pathlengths between col-
r 21~ lisions. If it is assumed that the problem is time-
~,~(,,z,~)= [ I l d,a(r' ,c~' )e - p ~x +~(xp+ ) - , dependent, each photon history is started by
00 assigning a set of values to the photon, its initial
energy, position and direction. Following this, the
g o r o 21
number of mean free paths that the photon propa-
x z2r'ddp'dr'+ ~ ~ I Sa(r',z',ck')
0 0 0
gates is determined stochastically. Then, the cross-
section (or absorption and scattering coefficients)
x flae-Pa~r~xp) - a(z-z')r'dc~'dr'dz' (4.4) data are sampled, and it is determined whether the
collided photon is absorbed or scattered by the gas
r 2x
molecules or particles in the medium. If it is
Fo,,(r,z,)= [ ~ ld,a(r',')e-~,~;(x;) -" absorbed, the history is terminated. If it is scattered,
00
the distribution of scattering angles is sampled and a
z o r o 21t new direction is assigned to the photon. In the case
x r'sin(~b- ~')zr'ddp'dr'+ ~ J ~ Sa(r',z',q~') of elastic scattering, a new energy is determined by
0 0 0 conservation of energy and momentum. With the
new set of assigned energy, position, and direction
x flze-P*%(x~)- ar'sin(q~-dp')]r'dO'dr'dz '. (4.5)
the procedure is repeated for successive collisions
until the photon is absorbed or escapes from the
When ld,a is interpreted as the wall function which
system.
includes the diffuse emission and reflection from the
Monte Carlo calculations yield answers that
walls, the additional integral terms will appear on the
fluctuate around the "'real" answer. As the number of
right-hand-side of these equations. It should be noted
photons initiated from each surface and/or volume
that in deriving these expressions, the medium is
element increases this method is expected to con-
assumed to be homogeneous. The evaluation of the
verge to the exact solution of the problem. Since the
integrals in these equations yields exact results for
directions of the photons are obtained from a
the radiative flux distributions in the medium. These
random number generator, the method is always
equations can be integrated numerically, as closed
subject to statistical errors and the lack of guaranteed
form solutions are not possible unless further simpli-
convergence. ~46 However, as next generation com-
fications are introduced. Considering that in most
puters become more readily available, Monte Carlo
engineering systems the medium is inhomogeneous
methods are expected to become more attractive for
and radiative properties are spectral in nature, it can
engineering applications. It has already been shown
be concluded that the exact solutions for RTE are not
that vectorization of the Monte Carlo computer code
practical for engineering applications. Nevertheless,
yields significant improvements in efficiency using
exact solutions for simple geometries and systems are
supercomputers such as CYBER-205 and more
needed, as they can serve as benchmarks against
precise results are obtained. 1s o
which the accuracy of other approximate solutions
are checked.

4.3. Zonal Method


4.2. Statistical Methods
The zonal method, which is usually known as
The purely statistical methods, such as the Monte Hottel's zonal method, T M is one of the most widely
Carlo method, usually yield radiation heat transfer used methods for calculating radiation heat transfer
predictions as accurate as the exact methods. The in practical engineering systems. In this method, the
124 R. VISKANTA and M. P. MENG0~

surface and the volume of the enclosure is divided As originally formulated TM the zonal method has
into a number of zones, each assumed to have a some inherent limitations, such as the treatment of
uniform distribution of temperature and radiative non-gray, temperature dependent radiative proper-
properties. Then, the direct exchange areas (factors) ties of combustion gases. The effects of temperature,
between the surface and volume dements are pressure and different species on gas properties can
evaluated and the total exchange areas are deter- be accounted for by weighted sum-of-gray-gases
mined using matrix inversion techniques. For an m o d d g 5~'5z In addition, it is usually difficult to
absorbing and emitting medium, the calculation of couple the zonal method with the flow field and
direct exchange areas becomes complicated as the energy equations which are usually solved using
attenuation of radiation along the path connecting finite difference or finite element techniques. This is
two area (area-volume and volume--volume) elements mainly because of the different size of the control
must be taken into account. volumes required; the zonal method can be comp-
The zonal method reduces the radiative transfer utationally prohibitive if the same grid scheme used
problem to the solution of a set of nonlinear by the finite difference equations is adopted Steward
algebraic equations. The set of energy balances for and Tennankore ~57 have coupled the zonal method
the zones in a closed radiaton system is written as with finite difference equations in modeling a
combustor by adapting two different grid schemes;
SE = Q (4.6) one for the radiation part and the other for the flow
and temperature fields. Recently, Smith et al. ~54 have
where combined the zone method with momentum and
energy equations to predict heat transfer in an
-Xs, S2St ... S,,Sl absorbing, emitting, and scattering medium flowing
J in a cylindrical duct.
The zonal method can not be readily adopted for
Sj S2 S2S2 - ~ S 2 S j .. S,,S2 problems having complicated geometries, since
J
numerous exchange factors between the zones must
S= S~ S3 S2S 3 ... SnS 3 be evaluated and stored in the computer memory.
However, this difficulty can be overcome by adapting
a hybrid solution scheme which employs both zonal
and Monte Carlo methods. This will be discussed in
SLS,, s,s ... s,s,-Es,,s Subsection 4.5.5 Note that the direct-exchange areas
J for rectangular enclosures have been recently calcu-
lated by Siddal115a who employed a new approach
12hl I Qt [ for the evaluation of the multiple integrals With this
I.~h2 ] Q~ I technique, it is possible to obtain these factors with
E= Eh3 and Q= Q3I any degree of accuracy desired It is worth noting
that the computer time required by the zonal method
in predicting radiative transfer in enclosures is
usually smaller than the time required by its
.Elm. .Q.J alternatives, and therefore the method is attractive
for practical engineering caiculations.l 56

Here, SiS~ is the total exchange area which is the


4.4. Flux Methods
ratio of the radiant energy emitted by a zone Si
which is absorbed by zone S~ (directly or after The radiation intensity is a function of the
multiple reflections from other zones) to the total location, the direction of propagation of radiation
energy emitted by zone S, Eu is the total blackbody and of wavelength Usually the angular dependence
emitted flux and Q~ is the imposed heat flux at zone of the intensity complicates the problem since all
Si.152 possible directions must be taken into account. It is,
Although the formulation of the zonal method for therefore, desirable to separate the angular (direct-
an absorbing, emitting, and scattering medium has io.nal) dependence of the intensity from its spatial
been available 153 for a long time, it has been only dependence to simplify the governing equationg If it
recently applied to the solution of radiation heat is assumed that the intensity is uniform on given
transfer problems in a system containing scattering intervals of the solid angle, then the radiative transfer
particles. 154 Larsen and Howell 155 presented an equation can be significantly simplified as the
alternative formulation of the zonal method and integrodifferential RTE equation would be reduced
accounted for only the isotropic out-scattering from to a series of coupled linear differential equations in
each volume element. This new approach, however, terms of average radiation intensities or fluxes. This
does not show any computational advantage over the procedure yields the flux methods. By changing the
classical zonal method.~ 56 number of solid angles over which radiative intensity
Radiation heat transfer 125

is assumed constant, one can obtain different flux an absorbing, emitting, and scattering medium are
methods, such as two-flux, four-flux or six-flux comprised of six coupled partial differential equ-
methods. Intuitively, one can deduce that as the ations. 165 The equations are quite complex and
number of fluxes increases the accuracy of the lengthy; therefore, they are not given here.
method would increase. Indeed, if the number of In general, the accuracy of the flux approximation
solid angles and corresponding directions are deter- depends on the choice of solid-angle subdivisions. If
mined from basic mathematical principles (see, e.g. there is no intersection between two adjacent sub-
Whitney 159) more accurate and efficient flux divisions, more accurate results are expected. 165 This
methods can be warranted. It is also possible to use has also been observed by Selcuk and Siddall a66 for
non-uniform solid angle divisions in the spherical rectangular enclosures. If the distribution of radi-
space. For example, if the direction and size of the ation intensity is assumed for each subdivision, the
solid angles are determined from the Gaussian or general equations given by Abramzon and Lisin t65
Lobatto quadratures, a non-uniform flux approxi- can be simplified and solved simultaneously. If the
mation is developed and the resulting expressions are fluxes in each subdivision are assumed constant, a
called the discrete ordinates approximation to the simpler six-flux model can be obtained from the
RTE. 15 general flux equations. For an absorbing, emitting
Another way of avoiding complicated expressions and scattering medium, Spalding ~67 suggested a
of the RTE due to the angular dependence of the similar six flux model for cylindrical geometry, which
intensity is to integrate the radiative transfer is written as
equation over the space after first multiplying it by
certain directional cosines. The resulting expressions ld
are called moment approximations. The spherical + r drr [rJ~a] = -(xa + crx)~a + xaEoa(T)
harmonics approximation is developed similarly, but
a more elegant and mathematically sound method of (4.7)
r
integration of RTE is employed. If the integrations
are performed over hemispheres or quarter-spheres,
then double or quadruple spherical harmonics - d~ (K~) = - (xa + o'a)K:~ + xj.Eba(T)
approximations are obtained, respectively. The first
order moment, spherical harmonics, and first-order
+ 6 ( J ~ " +J~- +K~" +K~- +L~" +Lj-) (4.8)
discrete ordinate methods are identical for the one-
dimensional, planar geometry; 16 however, they
differ from each other slightly for multidimensional 1 d
geometries. -+ r cl~ (L~) = - (xa + (ra)L~ + xaEba(T)
Due to the simplicity of the governing equations,
several flux methods have been developed for one- + ~ ( J ~ +J~- + K ~ + K j + L ~ + L j ) (4.9)
dimensional plane-parallel media. They are reviewed 0
elsewhere, T M and those which can be extended to
multidimensional geometries are compared against where J~, J j are spectral fluxes in positive and
experiments ~62 as well as against exact solu- negative radial (r) directions; K~, K j- in positive and
tions. T M In this discussion, the focus is on negative axial (z) directions; L~, L~- in positive and
multidimensional models. negative angular (4)) directions. These equations can
be manipulated to obtain three second order differ-
ential equations:
4.4.1. Multiflux models
Ever since the publication of the pioneering works 1 dfl- r -Id + }
of Schuster (in 1905) and Sehwarzchild (in 1906) on
the two-flux approximation, as flux models have been
one of the most used methods for radiative heat = tCa[J~" + J f - 2 Eba(T)]
transfer calculations. With the advances in com-
puters, the extensions of flux models for the appli-
cation to multidimensional systems have become + ~ra [2(J~ + J ; ) - K ~ - K ~ - L~ - L ~ ] (4.10)
r
possible, and consequently several different versions
have been proposed over the years. Recently, 1 d
Abramzon and Lisin 165 have presented a general
analysis for flux models in a three-dimensional
rectangular enclosure and have shown that most
other models reported in the literature can be = xaEK~ + K ~ - 2Eba(T)]
obtained from this general formulation. The gover-
ning equations for the general flux approximation in +3[-J~ -J~- +2(K~" + K ~ ) - L ] - L ; ] (4.11)
a three-dimensional cylindrical enclosure containing
126 R. VISKANTAand M. P. MLmG09

ldfl- 1 -Id + } where, J,~, J~', K~ and K~- have the same meanings
as defined before.
The four unknown fluxes in Eqs (4.13)-(4.16) are
= x~[ L I + L~- - 2 E ~ ( r ) ] determined from the four equations, and then the
radiative fluxes and the divergence of the radiative
O% flux vector are obtained readily. This method was
+3 [ -JI - J ~ - - K ~ - K ~ - +2(L + +L~-)]. (4.12) used to predict non-gray radiation heat transfer in an
axisymmetric furnace and good accuracy was ob-
These are the simplest forms of the flux equations tained) 72 Note that, although scattering in the
and can be easily written for axisymmetric enclosures medium was neglected in deriving these expressions,
as a four-flux approximation. The derivation of these it can be accounted for in the formulation. Also,
expressions is based on the Schuster-Hamaker these equations can be modified to relax the
method ~63'~64 which is the crudest and the least axisymmetry assumption to obtain a more general
accurate flux approximation for one-dimensional formulation.
systems. Whitacre and McCann~6S showed that the One of the oldest multi-flux methods is the six-flux
four-flux version of this model t69 predicted the method of Chu and Churchill. 173 Although it was
temperature field accurately, whereas the radiation developed for a one-dimensional, plane-paralld
fluxes were usually underestimated in comparison to medium, it is possible to modify this method for
Hottel's zonal method. A close examination of Eqs multidimensional enclosures. Varma ~4 obtained a
(4.7) to (4.9) reveals that the fluxes for one direction four-flux model for axisymmetric cylindrical enclos-
are not coupled with those of the other directions if ures starting from this six-flux method. However, the
the medium is nonscattering. A similar type, un- comparisons with more accurate models show that
coupled four-flux model was also developed by this version of the four flux method is not very
Richter and Quack IT and applied to a pulverized reliable. 17s Note that both the four- and six-flux
coal-fired furnace. methods account for the scattering of radiation.
In one-dimensional systems, the Schuster- Another six-flux model was proposed for three-
Schwarzchild two-flux approximation or its modified dimensional enclosures containing absorbing and
form ~@*'17~ yields more accurate results. Lowes et emitting gases. ~e6 A comparison of the predictions
al) 72 extended this method to axisymmetric en- based on this model with the Monte Carlo results
closures and derived an alternative four-flux model. showed that the maximum error in the radiation heat
The corresponding equations can also be obtained flux was not more than 23 % and could be reduced to
from the general relations by assuming axial sym- about 1% if the subdivisions of the solid angles were
metry and defining the boundaries for the sub- adjusted according to the geometry of the furnace.
divisions, t65 Then the governing equations* be- There are mainly three objections to the multi-flux
come 1 7 2 approximations of the radiative transfer equation
developed and used by some investigators for
d + (JI - practical problems (see Smoot and Smith 3 and
2 drr [J~ - J~ ] 4 r Khalil 5 for extensive lists of references and appli-
cations). First, there may be no coupling between the
axial and radial fluxes, which makes the equations
= -Ttxa(J~ +J~)+2xaEba(T ) (4.13)
physically unrealistic. Second, the approximation of
the intensity distribution from which the flux
~/~-~ d + ~ / ~ (J~ -J~ - r ~ -K~-) equations are obtained is arbitrary. Third, the model
2 dr [ J z + J ~ - ] 4 4 r equations cannot approximate highly anisotropic
scattering correctly, although it is theoretically
possible.
= -nxz(J~" + J~') (4.14)

x/~n d + x/~S-~n (J; - J~) 4.4.2. Moment methods


2 dzEK~ - K ; ] - ~ 4 r In the moment methods, the radiation intensity is
expressed as a series in products of angular and
= - n x x ( K f +Kf)+2xaEba(T) (4.15) spatial functions:

d I(x0,,z,0,+)
N
2 dz[K~+Kf]=--nxa(K~-K;) (4.16) (4.17)
= A o + ~ [~'A,.~+q"A,.,+II'A..j
n=l

*Note that these equations are modified slightly to follow where A's are functions of location only; ~,~/, and g
a consistent nomenclature. are direction cosines in x, y, and z-directions,
Radiation heat transfer 127

respectively [see Eq. (2.8)1. Although this equation is mation, is one of the most tedious and cumbersome
written in Cartesian coordinates, it can be given for of the radiative transfer approximations; however, it
any orthogonal system. As the upper limit of the may be the most elegant one because of its sound
series N approaches infinity, this expression con- mathematical foundation. The method was originally
verges to the exact solution for the radiation developed, as most other approximations, for study of
intensity. Note that Eq. (4.17) can be considered as radiative transfer in the atmosphere, 17s later modified
the Taylor series expansion of the intensity in terms for the solution of neutron transport problems, T M
of direction cosines. and extensively used for one-dimensional radiative
The simplest moment expression for the intensity transfer problems. 15-17Ag"2'za2'179 Although the
can be obtained by taking N = 1. This is called the formulation of the spherical harmonics approxi-
first-order moment method. The AI.~, A~.y and A~,.. mation for multidimensional geometries was dis-
coefficients can be obtained by integrating the cussed some time ago, m3t only during the last decade
intensity over the entire space. DeMarco and has the method been extended to two- and three-
Lockwood 176 have suggested some modifications of dimensional systems. For non-scattering Cartesian,
the moment method using the flux definitions of the cylindrical and spherical media the first-order (P~)
Schuster-Schwarzchild model, and defined the coef- and third-order (P3) spherical harmonics approxi-
ficients as mations,~ ao.~81 for an isotropically scattering cylin-
drical medium the PI -approximation, 182.184 and for
Ao=0 an isotropicaily scattering two-dimensional rect-
angular medium Pt- and P3-approximations ~ss
A~ ..,.=(J; -J;)/2, have been formulated and solved. Meanwhile, the
A~.,=(I<;-K;)/2, A~.:=(L~-L;)/2 (4.18) first-order spherical harmonics approximation has
A2.x=(J~ + J~-)/2, also been formulated to study the effect of cuboidal
A2.r=(K~ +K~-)/2, A2.:=(L~" +L~-)/2 clouds on radiative transfer in the atmosphere. 144't s6
Most recently, Menguc and Viskanta 61'187 reported
where A's are implicit functions of the wavelengths of the general formulations of the PI- and P3-
radiation 2, and J**, K,a , La + are integrated approximations for absorbing, emitting, and aniso-
spectral radiation intensities over appropriate solid tropically scattering medium in two-dimensional,
angles in the _+x, +y, +z-directions, respectively. finite cylindrical as well as three-dimensional rect-
These equations were solved by dividing the total angular enclosures.
solid angle 4n into six equal angles of 4n/6, each one In the spherical harmonics approximation, the
having the coordinate directions as its symmetry radiation intensity is expressed by a series of
axis. Another solution scheme was also adopted by spherical harmonics instead of a Taylor series and is
choosing a magnitude of 2n for each solid angle. .76 written as t a2
Although the latter assumption produces overlapping
of the solid angles, the predictions based on it yielded
better agreement with the Monte Carlo results for a t~(x,y,z,O,) = F~ A~Ax,y,z)r~(O,)(4.20)
n~O m = --n
three-dimensional rectangular enclosure. ~76 A further
improvement of this method was recommended by with
allowing some flexibility in the magnitude of solid
angle corresponding to each direction. 177 For a r.~(O,O)= ( - 1 ~" I,.j~/2
medium with a minimum optical thickness (ab-
sorption coefficient-characteristic length product) of r2.+ 1 (.-Iml)!-I
x _ ,~..,./
'/=.j.j, ..
r , ~cosvle
,.,,
(4.21)
2 this modified method yielded more accurate results L 4= ~,,+lml)!j
in comparison to the earlier versions. In neither of
these models ~76.177 was scattering of radiation in the where Y~ are the spherical harmonics, and P~ are the
medium accounted for. It is interesting to note that if associated Legendre polynomials which are related
the A-coefficients of this formulation are approxi- to the Legendre polynomials.
mated as In Eq. (4.20} the upper limit N for the index n is
known as the order of the approximation. Exact
A2.x = A2..v= As,". (4.19) solution of the RTE is obtained if N is taken as
infinity; however, for practical calculations a finite N
then the first-order moment method will be obtained value is assigned. N = I results in P1- and N=3
(as ~2+r/2+/t2= 1), which is equivalent to the first- results in P3-approximations. Usually, the odd
order spherical harmonics P~-approximation. 19 orders of spherical harmonics approximation are
employed, although there are occasionally some
others which use even order approximations, lsa The
4.4.3. Spherical harmonics approximation
reason for using the odd-order approximation is
The spherical harmonics (Ps) approximation, simply to avoid the mathematical singularity of the
which is also known as the differential approxi- intensity at directions parallel to the boundaries. The
128 R. VISKANTAand M. P. MENGOt;

radiation intensity is usually discontinuous at the radiation fields, but at the expense of additional
interfaces; therefore, it is not possible to have a single computational effort. It is shown 61 that the accuracy
value of intensity at the boundary. Consequently, it is of P3- as well as Pl-approximations can be sub-
not desirable to have an angular grid point just on stantially improved by using "exact" boundary
the interface. The roots of Legendre polynomials conditions, rather than somewhat arbitrarily defined
used in spherical-harmonics approximation yield Mark's or Marshak's boundary conditions (see Refs
Gaussian quadrature points, where the N-th order 19, 20, 131 and 132 for definitions and 61, 185 and
polynomial gives the N-th order Gaussian quad- 187 for implementations of the Marshak's boundary
rature scheme. If N is even, one of the quadrature conditions).
points will have a value of zero, which corresponds It is also possible to improve the accuracy of the
to an angular grid point on the boundary, whereas, if spherical harmonics approximation by obtaining the
N is odd there will be no quadrature point on the moments of radiation intensity in half or quarter
boundary. Therefore, an odd-order spherical har- spheres, xsg-19a Since the angular variation of
monics approximation yields a more stable solution. moments is allowed for in this method, the aniso-
The above discussion can be easily followed for a tropy of the radiation field can be modeled more
plane-parallel geometry. accurately than by the Pt-approximation. On the
The Pt-approximation is comprised of a single other hand, the governing equations are simpler than
elliptic partial differential equation ~a7 those for the Pa-approximation.

V210,a = Aa[10,a-4nlh4(T)] (4.22)


4.4.4. Discrete-ordinate approximation
where 10.4 is the spectral zeroth-order moment of A discrete-ordinate approximation to the radiative
intensity I-irradiance cga, see Eq. (2.21b)1, lb4 is transfer equation is obtained, as the name suggests,
Planck's blackbody function, and A4 is the coefficient by discretizing the entire solid angle (f2=4n) using a
which is a function of single scattering albedo 094, finite number of ordinate directions and corres-
extinction coefficient il4, and phase function para- ponding weight factors. The RTE is written for each
metersJa and ga: ordinate and the integral terms are replaced by a
quadrature summed over each ordinate. Originally
10.4 = S ladle, suggested by Chandrasekhar 15 for astrophysical
n=4x
problems, the discrete-ordinates method has been
A 4 = 3fl](1 - 094)[1 - o~4(ja + 04 -J]94)'] (4.23) extensively applied to the problems of neutron
transport. 21`13a'lq'*A95 A simpler version of this
In writing the above approximation, the delta- method, which is called SN-approximation, was
Eddington phase function is employed [see Eq. obtained by dividing the spherical space into N
(3.24)-]. In the P3-approximation, higher order equal solid angles, a96 However, more accurate form-
moments of intensity, i.e. the integrals of radiation ulations were obtained later using Gaussian or
intensity-direction cosine products over all direc- Lobatto quadratures. These are also called SN-
tions within solid angle 4n are employed. Naturally, approximations to symbolize the discrete-ordinates
the resulting equations are more complicated than approximation in which there are N discrete values
those of the Pt-approximation. For axisymmetric of direction cosines ~., q., it., which always satisfy the
cylindrical geometry, there are four second order identity ~.2 + q.2 +/~.2 = 1.
elliptic partial differential equations for the P3- In one-dimensional plane-parallel media, the discrete
approximation; 6~ whereas, for three-dimensional ordinates approximation has found many appli-
rectangular enclosures six equations are needed, ls7 cations (see, e.g. Viskanta, a34 Houf and Incropera, 197
These equations are solved simultaneously for the Khalil et al.t9s). Recently, the SN-approximation has
second-order moments, and afterwards the other been applied to two-dimensional cylindrical and
moments, radiation intensity, radiation heat fluxes rectangular radiative transfer problems with combus-
and the divergence of radiation heat flux are tion chamber applications in mind, and reasonably
calculated. accurate results were obtained in comparison to
Although the P~-approximation is very accurate if exact solutions? 75,199
the optical dimension (i.e. the product of extinction The radiative transfer equation for an axisym-
coefficient and characteristic length) of the medium is metric cylindrical enclosure is written for each
large (i.e. greater than 2), it yields inaccurate results quadrature point n as
for thinner media, especially near the boundaries.
Also, if the radiation field is anisotropic, i.e. there are
large temperature and/or particle concentration r Or r ~dp q-p,~-;+fl4la.,
gradients in the medium, the P~-approximation
becomes less reliable. The P3-approximation, how-
0" 2
ever, can yield accurate results for an optical =xalb4+7~,w., ~.,.14,.. (4.24)
dimension as small as 0.5 61.~a5 and for anisotropic t/l: n"
Radiation heat transfer 129

where w, is the weight of the Gaussian quadrature Here, we discuss only those which are applicable to
points. Integrating Eq. (4.24) over an arbitrary combustion problems.
control volume and rearranging yields, The basic flaw of the zonal method is the
computational effort required to calculate the ex-
change factors between various volume and surface
{ ~.(ANI a,.,N -- Asla,.,s) + 14,(ArJ a..,v. - Awl~,~,w) elements in complex geometries. This difficulty can
be overcome using the Monte Carlo method to
1 calculate the direct exchange areas. Is2 If the radi-
- (As - As)-- (~. + 1/2I~.~ + 1/2,c
Wn ative properties of the medium are known and do not
depend on temperature, it is possible to calculate
-~,-1/2I~., - 1/2,c)}/Vc these exchange factors only once and store them in
the memory of a host computer for later use in the
zonal method predictions. By doing this, the compu-
0",1
= - flalx.n.c + xxlba.c + 7 - ~, Wnn' ~,m'lx,,',C (4.25) tational time required by the zonal method to predict
t'l'Tt n'
radiation heat transfer in complex geometries is
decreased substantially. However, the computer stor-
where A is the corresponding area of control-volume age requirements can become prohibitive if the
side for N, S, E or W, i.e. for north, south, east, or number of zones is large.
west side, respectively; V is the volume of the control The Monte Carlo method suffers from statistical
volume, C is for the central node, and or-terms are to error as well as the extensive computational time
preserve the conservation of intensity in the curved required for the calculations. If the direction of each
coordinate, which are determined from the radiative ray is given deterministically rather than statistically
equilibrium condition.~ 99 These governing equations and if all the directions constitute an orthogonal set,
are solved numerically, for example, using a finite- then the solution would be less time-consuming and
difference scheme. 175'199 A finite element solution the accuracy would increase with the increase in the
scheme was also developed to solve the discrete number of directions. With this in mind, Lockwood
ordinates approximation equations in two-dimensional a n d S h a h 23'24 proposed a "'discrete transfer"
Cartesian geometry for radiative transfer in the model which combines the virtues of the zonal,
atmosphere. 2 Monte Carlo, and discrete ordinates methods. They
If the resulting equations of the discrete ordinates showed that very accurate results could be obtained
approximation are carefully coded, they can result in with this method in one- and two-dimensional
computer algorithms that combine minimum com- geometries by increasing the number of directions.
puter memory requirements with few arithmetic
Although this method is claimed to be capable of
operations per space-angle grid point. 133 However,
accounting for scattering in the medium, no results
this approximation is not flawless, but suffers from have been reported or compared against other
the so called "ray effects" which yield anomalies in benchmark methods for scattering media in multi-
the scalar flux distributiori. T M .202 The ray effects are dimensional enclosures. The results for a one-
especially pronounced if there are localized radiation
dimensional scattering medium did not show the
sources in the medium and scattering is less im- same level of agreement with the benchmark results
portant in comparison to absorption. As the single as did the non-scattering medium predictions. 23
scattering albedo increases, the radiation field be- This method is also likely to yield erroneous results
comes more isotropic and the ray effects become less
due to the "ray effects" discussed in Subsection 4.4.4.
noticeable. However, with increasing single scat-
A similar approach to the solution of the RTE for
tering albedo and/or optical thickness of the
multidimensional enclosures has also been presented
medium, the convergence rate may become very by Taniguchi et al. 2s for absorbing-emitting media.
slow. 133 Considering the flame in combustion cham-
This so called "'radiant heat ray method" is based on
bers as a localized radiation source, it is natural to
the Beer-Lambert's or Bouguer's law and yields the
anticipate the ray effects in the solution of the RTE
radiant energy absorption distribution in noniso-
in combustion chambers, if the discrete-ordinates thermal enclosures containing combustion gases.
approximation is used. If the combustion chamber is
Comparisons of the predictions based on this
a pulverized-coal fired furnace in which there are
method with other results show that the method is
scattering particles present, the results are expected
more accurate and less time consuming than both
to be more reliable. 199
zonal and Monte-Carlo techniques if the radiative
properties such as the absorption coefficient and wall
4.4.5. Hybrid and other methods emissivities are constant. 2os
Another hybrid model based on the Monte Carlo
Almost all methods discussed have some flaws. In method and generalized radiosity-irradiation ap-
order to take advantage of the desirable features of proach has been suggested by Edwards. 26 This
the different models, various hybrid radiative trans- method accounts for the volumetric scattering, yields
fer models have been developed in the literature. accurate results for optical dimensions as small as
JPBCS 1 3 : 2 - e
130 R. VISKANTAand M. P. M~,~GOt;

0.5, and is computationally faster than the Monte accuracy and computational costs. In order to decide
Carlo method. whether a model is appropriate for a given problem,
The main reason why the discrete ordinate one has to compare its predictions against the
approximation suffers from ray effects is because of benchmark results obtained from either experiments
the inability of the low-order Sn-quadrature to or exact solutions. Zonal and Monte Carlo methods
integrate accurately over the angular flux. 133 If are extensively used as the benchmark for compar-
piecewise continuous approximations of the angular isons as they generally yield accurate predictions of
flux are given in terms of directional variables, and radiation heat transfer.
approximate spatial equations are obtained by In one-dimensional systems, comparisons of differ-
integrating over appropriate solid angles, these ray ent radiation models have been
effects can be avoided. The resulting expressions can given. 125']62-164"197']98 However, the accuracy of a
be considered as hybrid models which combine method in predicting radiative transfer in a simple
discrete ordinates or multiflux approximations with system may not always warrant its use in more
the spherical-harmonics approximation. Indeed, the complicated systems. Therefore, it is important to
double or quadruple spherical-harmonics approxi- evaluate radiative transfer models for multidimen-
mations described in Subsection 4.4.3 can be con- sional geometries, preferably for practical situations.
sidered as this kind of hybrid model. In neutron
transport literature there were several accounts
which discussed the possibility of combining the Sn-
method with the Pn_~-method to improve the 1.0
accuracy and reliability of the predictions as well as FiniteElement
to decrease the computational effort, t 32.133
0.8~ L / H ,m o Zonal
Flux models can also be coupled with the moment
or spherical harmonics approximation to improve
the accuracy of the radiation heat transfer pre- 0 . 6 ~
dictions. A model which combines the Pt-approxi-
mation with a two-flux method was proposed by
O.4
Selcuk and Siddall 27 and applied to a two- ~o
dimensional axisymmetric cylindrical furnace. The
comparisons of the temperature and heat flux distri- i5
O.2
butions in the medium with those obtained with the
zonal method showed very good agreement. Since O | I I ~ I I
this model was developed for a gas-fired furnace, 0 0.2 0.4 0.6 0.8 1.0
scattering of radiation was not taken into account. DimensionlessPosition,x/L
Another similar hybrid method was derived by FiG. 13. Dimensionless centerline temperature profiles in
Harshvardhan et a/. 28 who combined the modified rectangular enclosures of different aspect ratio with black
two-flux method]7~ with the P~-approximation. In walls; bottom wall at dimensionless temperature 0= 1.0,
this method, the linearly anisotropic scattering other walls at 0=0. 2~2
medium assumption was made, and the method was
used to predict radiative transfer through three- ~" ILlH .O.t
dimensional cuboidal clouds. Comparisons of the
predictions with the Monte Carlo results showed
..'--
reasonably good agreement. ,'7" 0.8 ~ - .... ~"*"*
Recently, a new three-dimensional radiative trans-
fer model was proposed 2o9.2~o by extending the one-
dimensional adding-doubling technique (see, e.g. van .o 0.6
de Hulst 2It). The predictions for the radiative flux "0

distribution in cuboidal clouds obtained by this


method compare very well with those of the Monte ZlU 0.4 FiniteElement
Carlo method. It should be noted, however, that the
---P3
_~ o Zonal
assumption of homogeneous and symmetric bound- .~_
ary conditions simolify these problems considerably. 0.2

4.5. Comparison of Methods I I !


Oo o'.2 a, ols o8 ,.o
Although there are several radiative transfer
models available, it is difficult to choose a "best"
DimensionlessPosition,x,/L
model for different applications. For a given physical FIG. 14. Dimensionless net radiation heat flux at the lower
wall in rectangular enclosures of different aspect ratio with
situation, one of the several models can be used black walls, bottom wall at dimensionless temperature
according to the applicability of the model, desired 0= 1.0, other walls at 0=0. 2~2
Radiation heat transfer 131

2.8 accuracy of the surface net radiation heat flux

"" 2.4 . ~
.o

-'-Ps
---
Zonal

~
decreases with decreasing optical thickness. Similar
conclusions have also been reported by different
researchers.6~'l s5.1 s7
2.0 In Figs 15 and 16, comparisons between zonal,
,5 spherical harmonics (P3), and discrete ordinates (SN)
(J
- 1.6
approximations are presented for a purely scattering
o
medium with different wall emissivities. 199 The P3-
and S6-results for the centerline irradiance distri-
.~_ 1.2 :'"i" ~. . . . . . . . bution are in very good agreement with the zonal
method (see Fig. 15). The Ps-approximation, how-
,,, 0.8
e, ever, overestimates the radiation heat flux at the
.=_o walls for large wall emissivities, although both S4-
a4 and S6-approximations yield accurate results (Fig.
E 16).
0 The lower-order spherical harmonics approxi-
0 o', & & ,o mations generally yield more accurate predictions if
Dimensionless Position, y/H
the radiation field in the medium is almost isotropic,
FIG. 15. Comparison of irradiances in a two-dimensional which is the case if the optical thickness is large,
square cross-section enclosure with a gray scattering
medium (ic=0), Ehl = 1 and El,_,= EI,.~= El,4=0.199
the medium is predominantly scattering or the
surfaces are diffusely reflecting. If the radiation field
is highly anisotropic, the P3- and especially P t "
approximations become less reliable. Because of this,
Zonal the Pi- and P3-approximations are to be used for
1.0 ~'-,,, - - ' - - Ps media having optical thicknesses of 1.0 and 0.5
,~ S,, Ss
or larger, respectively, ls'lsSas7 The main reason
for this inaccuracy for anisotropic radiation fields
~ 08
h is use of arbitrarily defined boundary conditions,
like Marshak's condition) 9 In Fig. 17, the P3-
:I: approximation results are compared against those of
O.6
o an exact model 139 for a cylindrical enclosure, 61
where both Marshak's (m) and "exact" analytical (a)
~ O.4 .,,.::.~---;--~---~ ........ boundary conditions are used. Here, it is assumed
that there is a uniform, diffuse radiation source
incident on one of the end surfaces of a cylindrical

(,0.1
0.8
0 I I I I
0 (11 0.2 03 0.4 05
Dimensionless Position, x//L O.6

FIG. 16. Comparison of radiation heat fluxes at a wall of a qr


two-dimensional square cross-section gray enclosure with a
scattering medium (h=0), Eht = 1. Eh2 = E~,3= Et,4=0.199
(r --to)0.4
.-m
: a ..........
Exact
PI

0.2 ~... ------ PS


N~.
Unfortunately, this is not always possible because of 0
0 2 4 z/ro6_ 8 K)
the analytical or numerical difficulties.
The radiative equilibrium (heat transfer by radi-
ation alone) assumption yields the most simple case la - - / m.-~ -,'--t ....... : , "
for solving the radiative transfer equation in two- or o.,I-. .............N ................ :: / oo,o
three-dimensional enclosures. In Figs 13 ad 14 the qz r- . . . . . . . . . . . . qz
(z=O) _ La_~-A_m"'~ ~ (Z=Zo)
centerline dimensionless temperature and net radi- o u F----'~'_.-_-~-__-_.. . . . ~__.=.__ OOOS
ation heat flux distributions at one of the walls of a I
rectangular enclosure containing a gray, absorbing- ozl l , ; ......... i ..... "'-'1o.oo o
o oz o.4 o.6 as ,.o
emitting medium are compared for three different
,/to
methods. 212 The zonal and finite element methods
FIG. 17. Comparison of P,- and P.,-approximation results
are in good agreement with each other. The third
with exact benchmark solution: 1",)=0.1 m. # = l . 0 m - ' ,
order spherical harmonics (Ps) approximation yields co=0.5, T~,.=O, 8 , = 1.0 (m refers to Marshak's boundary
good results for the distribution; however, the conditions, a refers to analytical boundary conditions). 6~
132 R. VISKANTAand M. P. MENG0q

i I , i i

0 Meosured Values
P,C A p p r o x i m a t i o n

I I I I i I a I i
IO 2.0 3,0 40 50
z (ml

FIG. 18. Comparison of local radiation fluxes at the wall based on P.~-approximation results for a
combustion chamber with experimental data and discrete ordinates method: r. =0.45 m, :, = 5.1 m."~

enclosure containing a homogenous, absorbing and N


150
scattering medium. Radiative flux disribution on the
cylinrical walls (upper panel) and on the end walls
'\
(lower panel) are plotted from the results obtained x I00

~
.o
using different boundary condition models. 61 It is Lt. O.Im "l
clear from the figure that the use of Marshak's
boundary condition yields substantially higher local ,,.~,li~d~ ~- '1 cleat 2 q t a y

errors in radiation heat flux than the use of analytical


conditions. This suggests that with a careful and I I I I I I
more rigorous treatment of the boundary conditions, 0 I 2 3 4 5 6

even the very simple P~-approximation can be Furnoce Lenqth (m)


employed to predict the radiation heat transfer in FIG. 19. Comparison of predicted radiation heat fluxes
combustion chambers accurately. using the four flux model (lines) with different absorption
There are also some accounts in the literature coefficient formuhltions,1 v2 Points are zonal method results.
which compare different radiation models for prac-
tical systems, such as large scale furnaces where there
is a gradual temperature variation along the cham- obtained very good agreement between the experi-
ber.61.168,172.175.213 In these comparisons, radiation mental data and one-dimensional radiation models
is decoupled from the energy equation, since the using the measured radiative property data in the
temperature distribution as well as the radiative models. From these findings, one may conclude that
properties of the medium are assumed to be given. In for complicated systems such as combustion cham-
Fig. 18, the predicted radiation heat flux distribution bers, the accuracy of the radiative properties is as
along the cylindrical walls of a furnace is shown. 61 important, or even more so, than the accuracy of the
Both the $4- and P3-approximations are seen to be models. Additional sensitivity studies must be per-
accurate for a given "uniform" absorption coefficient formed for combustion chambers to determine which
value of 0.3 m -1 However, it is difficult (almost properties are the most important and under what
impossible) to assign a single "gray" absorption conditions. For example, Menguc and Viskanta 2~4
coefficient for the entire furnace. When x is changed have shown that the index of refraction of coal
from 0.3 m -~ to 0.35 m -~ (see Fig. 18) the P3- particles does not play a significant role in predicting
approximation yields improved agreement between the radiation heat flux distributions along the
the data and predictions. If the value of x were furnace walls. In their model the pulverized coal
changed to 0.25, the agreement would have been particles were assumed to be only in the flame zone,
poorer. The sensitivity of the results to the radiative and the predictions were obtained using the P3-
properties was also shown by Lowes e t aL 172 They approximation. On the other hand, Piccirelli et
compared the predictions obtained from zonal and al. 215 presented a similar analysis using the PI-
four-flux models for different absorption coefficients approximation for a one-dimensional cylindrical
against the experimental data obtained for a gas- system and showed that the complex index of
fired furnace. As seen from Fig. 19, the results are refraction was a very important parameter in
more sensitive to the radiative properties than to the predicting the emissivity and absorptivity accurately.
models. Indeed, in predicting the radiation flux These contradictory conclusions are not due to
distribution in a large furnace, Selcuk et a/. 213 different solution techniques, but basically result
Radiation heat transfer 133

from assumptions related to the radiation property External


distributions in the medium. In the former 2~4 the Radiotion
coal particles were assumed to be only in the flame
zone, whereas in the latter 2t5 the coal particles filled
the entire combustion chamber.
Thin
~. VOlOtlle r~loua,~ Flame

5. APPLICATIONSTO SIMPLE COMBUSTION SYSTEMS


To illustrate the coupling between radiation heat
transfer, combustion and other transport processes \. '/r2 . ' " ."x/,Caol.
we consider in this section several simple combustion Parttele
situations in which radiative transfer has been
accounted for. The emphasis in the discussion is on
the effects of radiation heat transfer. Since the FIG. 20. Schematic diagram of a spherical translucent and
physical situations considered are quite simple and radiating cloud model. 218
the systems are not large, the effects of radiation on
the results are expected to be small as the optical
dimensions characterizing the systems are also small. temperatures. The spherical transluscent and radi-
ating cloud model is identical to the problem studied
earlier. 219 This complex energy transfer situation for
5.1. Single-Droplet and Solid-Particle Combustion solid particle combustion is treated rigorously, and
Burning of a single-droplet of liquid fuel or of a influencing parameters are identified.
solid particle is a very simple system. Transport The model permits calculation of the steady-state
process and not chemical kinetics dominate the heat transfer rate when the particle surface temper-
combustion of fuel. This phenomenon has been ature, flame-sheet radius and temperature and other
studied extensively for many years and experimental environmental conditions are given. The optical
and theoretical accounts are available, s'2~6'2~7 The thickness has been found to be an important model
theory of single-droplet combustion is complicated parameter in calculating radiative transfer rates. A
by many factors, such as circulation in the droplet, fair amount of numerical computation was required
finite-rate chemistry in the diffusion flame that to obtain solutions. The model has been shown to be
surrounds the droplet, nonsteady accumulation of useful for the interpretation of muiticolor 22 and
fuel between the surface and the flame, etc. In recent two-color T M pyrometry for more accurate experi-
reviews of the theory these complications have been mental data reduction. Ultimately, a simplified
discussed.S.217 Radiative transfer in single-droplet version of the model could be incorporated into a
combustion has been ignored in most studies reported coal combustion model which explicitly includes
in the literature. 8 Recently, a model has been particle heat-up and devolatization rates. However,
developed to study coal particle behavior under for an application to a combustion system the model
simultaneous devolatilization and combustion in must be extended to account for the interaction
which transport of radiation in the volatile cloud between the burning particle and the surroundings
(radiatively participating medium) surrounding a which contain radiating gases, clouds of particles
coal particle has been accounted for. 218 and the system walls.
The spherical volatile cloud, enclosed by a thin A systematic investigation of the effects of thermal
flame sheet whose location is determined by diffusion- radiation on the combustion behavior of char
limited combustion, is modeled as a radiatively particles exposed to an oxygen environment has been
participating medium. The modeling concept is performed 222 using the general mathematical models
similar to that of liquid droplet combustion except developed by Sotirchos and Amundson. 223'224
that volatiles emitted by the coal particle form a Pseudo-steady computations have shown that porous
concentric luminous mantle (see Fig. 20) and radi- char particles reacting under radiative equilibrium
ative transfer is important in addition to convective conditions are found to present considerably lower
and conductive transport. The model describes the burning times (by more than one order of magnitude
heat transfer mechanisms between the particle, the in some cases) than their heat-radiating counter-
volatile cloud, the flame, and the external en- parts. 222
vironment. The analysis of combined conduction-
radiation heat transfer in a concentric sphere filled
5.2. Contribution oJ Radiation to Flame
with a radiatively participating medium first devel-
Wall-Quenchin 9 of Condensed Fuels
oped by Viskanta and Merriam 2~9 was used. The
absorbing, emitting and scattering medium was The understanding of extinction phenomena has
assumed to be confined between two gray, diffuse been greatly improved over the years, and recently
isothermal spheres kept at different but uniform several important mechanisms of extinction or
134 R. VISKANTAand M. P. MENGOt~

quenching phenomena have been proposed. 225.226 function of various thermophysical and radiative
Among those proposed are the stretching effect of the parameters such as the conduction-radiation ratio,
combustion zone, preferential diffusion, buoyancy optical thickness and heat generation intensity by
and heat losses. The effect of heat loss on quenching chemical reactions. A new dimensionless group, the
can be more pronounced in the presence of relatively modified Damk6hler number (ratio of dimensionless
cold boundaries, due to steep temperature gradients. heat source intensity to the conduction-radiation
The pyrolizing surface of condensed fuels is a parameter), which characterizes the relative strength
representative example of cold boundaries in a of heat generation to radiation transport, emerges
combustion situation of practical interest. In this as from the analysis. The quenching-layer thickness is
well as in many other studies on heat transfer in fires, determined primarily by the conduction effect near
the significance of thermal radiation has become the relatively cold surface. However, thermal radi-
increasingly recognized as radiation accounts for a ation is still the dominant mode of heat transfer
significant portion of heat losses. Its effect has been there. Numerical calculations have shown that the
shown to be considerable not only in large-scale and fraction of radiative heat flux at the fuel surface is
small-scale fires 227'229 but also in small diesel over 85 ~o of the total heat flux. Optical thicknesses
engines. 2a less than 0.5 show little influence on the quenching
Radiation blockage by soot layers between the distance, and more opaque systems yield shorter
flame and the fuel is considered to be an important quenching distances.
characteristic of fires. The radiation blockage effect Radiation blockage may also be desired in other
has been investigated and found to depend on the physical situations to avoid excessive temperatures at
type of fuel and size of fires. 227 Using experimentally the system boundaries. Siegel ~ls has systematically
obtained data, it is shown that for polymer fuels studied a one-dimensional system at high temper-
of polymethylmethacrylate (PMMA), polypropylene ature, with and without flow, to determine the
(PP), and polyoxymethylene (POM) no significant governing parameters needed to keep the walls at a
radiation blockage is present in moderate-scale fires; prescribed temperature range. Using an analytical
however, for sootier fuels such as polystyrene (PS), approach, he concluded that a dimensionless para-
radiation blockage has a considerable effect even in meter ME= TJ~CLxL/C2) and the ratio of suspension
small-scale fires. In Fig. 21, the effects of gas layer temperature to source temperature, TraIT,, were the
thickness and soot volume fraction on the blockage two important parameters. Here, fv is the soot
of radiation are shown graphically. volume fraction, CL is the ratio of mean beam length
The effect of thermal radiation and conduction on to layer thickness, x is the constant absorption
the cold-wall flame-quenching distance in the com- coefficient, and C2 is Planck's second radiation
bustion of condensed fuels has been studied using a constant. When M ~ 2 the soot (or suspension) layer
simple physical model. 227 In the analysis a steady- absorbs practically all the radiation incident on it,
state, no-flow condition, one-dimensional energy and when M ~ 0 . 2 half of the radiation is absorbed.
equation for the optically thin quenching layer is Although at the beginning the soot layer blocks the
solved employing the singular perturbation tech- radiation, the energy trapped in the layer raises its
nique. The quenching distance is obtained as a temperature. After a while, the layer begins to radiate
energy. This can be avoided using perforated walls
and introducing the cool seeded gas from many holes
1.0 along the surface at frequent intervals)~S A similar

/, approach can also be applied to combustion cham-


bers, like the liners of gas-turbine engines, to predict
0.8 E~,- 0.9
the amount of film cooling necessary. It is worth
noting the similarities between these results and
those obtained by Lee et al. 227
0.6 .... E.. o . g s ~

5.3. E.JJect of Radiation oll One-Dimensional Char

'7/' / /
~ 0.4 Flames
fr ~ In one-dimensional pulverized-char or coal flames,
two different flame types are recognized as "small" or
0., / _
"long". T M The "small" type flames can be modeled
qualitatively using a conduction-diffusion approxi-
mation, whereas for the "'long" type flames, radiation
10-2 10"1 heat transfer is also an important mechanism. Earlier
GAS LAYER THICKNESS ( m ) attempts to model these types of flames without
FIG. 21. Radiation blockage as a function of gas layer including radiation have not been very successful;
thickness for a plane flame layer model, Lj./Lo=0.4. 227 however, a model based primarily on radiation
Radiation heat transfer 135

predicted the flame temperatures and burnout pro- the density of coal, and Q is the correction factor
files very accurately. TM In this model it is assumed which is near 1.5 for practical flames. Equation (5.2)
that reaction is controlled by combined diffusion and is applicable if a particle size distribution is given. It
surface chemical reaction for either shrinking, con- can be used also if a single mean value for "r" can be
stant density particles or constant diameter, de- defined. Two different mean values, one from the
creasing density particles. Also, the size distribution Sauter-mean diameter definition (i.e. volume-to-
of the spherical particles in the flame is accounted surface ratio, D32 ) and the other from the Rosin-
for; however, the particle temperature is assumed to Rammler index were also used in the analysis. The
be equal to that of the surrounding gases. This predictions for coal burn-out are compared against
approximation can not be justified in physical experimental data in Fig. 22 for polysize as well as
systems where particles burn in dilute suspensions two mean diameter models. The polysize model is in
with a large excess of oxygen, yet it is a reasonable exceptionally good agreement with the data, sug-
approximation for concentrated suspensions in prac- gesting that the accuracy of properties used in a
tical flames (see comments of I. W. Smith to Xieu et radiation model are as critical or, maybe, even more
al.231). The radiation heat flux was obtained by so than the accuracy of the model itself (see also
modeling the RTE between two vertical infinite Section 4.5).
parallel-plates, and the flux divergence is given by 23~ Another model, based on the one-dimensional
Eddington (P1) approximation for predicting the
c3q, contribution of radiation in "long" flames was given
.... 4xtr T4( z ) + 2x[ a T41E2( z )
dz by Krezinski et al. 233 They obtained the radiative
properties of coal particles using the complex
tL
refractive index data of Foster and Howarth; v5
+trT4.2E2('rl,--z) +tr S Ta(t)El(lz-tl) dt] (5.1)
o however, they did not compare model predictions
with experimental data.
where x is the absorption coefficient for the mixture,
T is the optical distance, tr is the Stefan-Boitzmann 5.4. Radiation in a Combusting Boundary Layer Along
constant, El and E 2 are the first- and second-order a Vertical Wall
exponential integral functions, respectively. The key
parameter in this model is the absorption coefficient, Classical studies of boundary layer diffusion
which, in general, is a function of location. The flames have neglected radiation, a.234.235 To isolate
assumption of a constant value for 1( did not yield the effects of radiation in flames from the complex-
accurate predictions for the temperature and burnout ities of fluid motion and to gain better understanding
profiles. 232 Recognizing that the absorption coef- of radiation heat transfer in fires, analyses have been
ficient is varying with the projected cross-section of made of a laminar, combusting boundary layer along
the particles (see Section 3.2), Xieu et al. TM used a a vertical wall. 236-23s The rate of upward flame
new expression spread over a vertical combustible surface is an
important parameter in the ranking of the fire hazard
~, = 3QD,/4pr (5.2) offered by different materials. Typically, when the
flame height reaches about 2 m, the flow becomes
where D, is the dust cloud concentration for the size turbulent and radiation heat transfer starts to play an
interval corresponding to mean particle radius r, p is important role in the overall energy balance. Free,
mixed and forced convection boundary layers along
a vertical, burning wall have been studied analytic-
ally. Previous work on thermal radiation from flames
'1 I i
has been reviewed 239-24~ and related experimental
work has also been reported. 242 Here, we discuss the
J[ / Rosin -Rommter monosize results of numerical solutions obtained for a steady,
laminar, radiating, combusting, boundary layer over
ca 60 "1 ~ C a vertical pyrolizing fuel slab.
An analysis has been developed for steady free and
40-1 ~ ~ % forced laminar combusting boundary layers in which
I \ I ' ~ Polysize
a pyrolysis zone separates the flame from the fuel
surface 23s as shown in Fig. 23. The soot layer is on
~- I monosize (D~l~" ~-.....__--_.~ the fuel side of the flame zone. Through the
transparent gas the combusting layer exchanges
I I I I I I I radiation with a distant black wall, which is
0 20 40 60 80 ioo J20
Distance from tube bonk (cm) maintained at a specified temperature. The chemical
FIG. 22. Comparison of prediction and experiment for coal energy lost to the system in the formation of soot is
burn-out, using the given polysize and two alternative, neglected. The effects of radiation on the local fields,
mort osi ze models. 231 and excess pyrolyzate escaping downstream at the
136 R. V1SKANTAand M. P. MENO0~

Here ~=xy and T6=x6 , where 3 is the boundary


thickness. In writing Eq. (5.5) a one-dimensional
radiative transfer model is used which is consistent
/ with the boundary-layer approximations.
/ I ecu~o~v Numerical solutions were reported for both forced
/ LAYER EDGE and free flow along a vertical pyrolizing fuel slab. 23s
The optically-thin approximation was assumed to be
valid for radiation. In the analysis of a combusting
/ boundary layer with radiation, the pyrolysis rate was
/ found to depend on nine dimensionless parameters
and the intensity of the external radiation flux. The
/ dimensionless heat of combustion (Dc) plays a
FUEL / dominant role in determining the flame temperature
/ AME ZONE and makes it a significant parameter in radiating
/ systems. In addition, the optical thickness of the
/ boundary layer and the radiation parameter (Na)
/ affect the emission from the combusting boundary
/ layer. The surface temperature and the emissivity
/ AMBIENT AIR characterize the surface emission, which can domin-
ate the flame radiation in the boundary layer for
solid fuels of small dimension. A comparison bet-
ween numerical and experimental pyrolysis rates
shows good agreement for a case where surface
emission dominates flame radiation, i.e. burning of
FIG. 23. Schematic of a steady, two-dimensional, laminar. polymethylmethacrylate (PMMA) in air. Values of a
radiating combusting boundary layer flame with soot on a mean absorption coefficient and soot generation
pyrolizing fuel slab.238
rates were also obtained using the analysis. This type
of data could be used to quantify soot formation
top of the fuel slab are examined by assuming that models. 23s
the dominant effect of the soot particles is on the
radiation heat transfer.
The conservation equations for mass, momentum 5.5. Interaction of Convection-Radiation in a Laminar
and species for a radiating-combusting boundary Diffusion Flame
layer are identical to those for a nonradiating
boundary layer s and therefore are not repeated here. Most of the earlier and even some recent studies
The energy equation is given as 23s dealing with high-temperature situations such as
those found in laminar diffusion flames have excluded
the effects of radiation (e.g. Refs 234, 235, 243-245).
{ Oh t3h\ ~ {k --I t3h\-dq'+S
p / u -- + v--1 = - - / - - (5.3) In many situations radiation from the hot gases can
significantly alter temperature in both the flame itself
and in the surrounding regions as well as within the
where the enthalpy is defined as flame structure. The relatively simple character of
7"
diffusion flames in laminar stagnation-point flows
I1= f cpdT. (5.4) has led to several theoretical and experimental
T studies of that system in which thermal radiation has
been incorporated in the analysis. 246- 249
Assuming a spectraily gray, homogeneous medium Interaction of convection and radiation on the
with a constant absorption coefficient, the local temperature and species concentration distributions
radiative flux in the y-direction* can be expressed in a diffusion flame located in the lower stagnation
as 23s
region of a porous horizontal cylinder 246 and a
vertical flat plate 249 have been studied experiment-
t
ally and theoretically. The exponential wide-band gas
q,=2[Eb~.E3(z)-E ~ E 3 ( 17~ - - T)--J'- SEb(t)Ez(z -- t)dt
radiation model was employed in this inhomo-
0
ta
geneous (nonuniform temperature and composition)
-- I Eb(t)E2(t - z)dt]. (5.5) problem through the use of scaling techniques. Using
a numerical scheme, the compressible energy, flow,
and species-diffusion equations were solved simul-
*There is an error in the expression for the radiation flux taneously with and without the radiative component.
given in Ref. 238, but this does not affect the validity of the In the experiment, methane was blown uniformly
results obtained because of the approximations. from the surface of the porous cylinder, setting up
Radiation heat transfer 137

Z.O ! I | absorption coefficient it is possible to match experi-


mental data with predictions; however, this approach
0 F 2600 c c / m l n is not based on first principles and requires data for
2.4 each set of conditions.
I[xponentlol Wide-Bond Ido~l The experimentally measured convective heat fluxes
- - m - - Groy GaS Model K'p-0.15 at the wall of the cylinder were found to be in better
~-~ No Rodiotlve Interoction
2.0 agreement with the results calculated using the wide-
Q I[xporimento! Run I
band than the gray-gas model. 246 This further
O l[aoerimental Run 2
supports the performance of the nongray model and
I.G shows that the model is superior to those based on
the gray-gas model as well as the results that ignore
f,.
I the effects of radiation. Measurements of convective
~ t 1.2 and radiative heat fluxes in a diffusion flame
surrounding a porous cylinder burning drops of n-
heptane have shown that radiation heat transfer to
3 the cylinder is by no means negligible. 24s Radiation
~. 0.8
accounts for about 40 % of the total heat transferred
to the cylinder, but the radiation from gases (CO2
0.4 and H 2 0 ) is only 20 % of the total radiation, with the
rest being soot radiation. For those types of flames
where soot radiation is more important than gaseous
0 I ~- radiation, the use of a gray model would yield
0 I.O 2.0 3.0 4.0 5.0 reasonable results.
II A similarity solution for an opposed laminar
diffusion flame with radiation has also been ob-
FIG. 24. Comparison of theoretically predicted and experi-
mentally measured temperature protiles during methane tained. 2'.8 In this combustion system a stream of
combustion around a horizontal porous cylindrical burner.2'.6 oxidizer approaches the stagnation point on a
condensed surface and reacts with pyrolyzed fuel in a
thin diffusion flame with a constant-thickness boun-
(upon ignition) a diffusion flame within the free- dary layer. The fuel surface is assumed gray and
convection boundary layer. Using a Mach-Zehnder diffuse, and the gas is considered gray. Only the
interferometer and a gas chromatograph, temper- pyrolysis region is considered. Numerical results
ature and composition measurements were obtained were obtained using the exponential kernel and
along the stagnation line. Excellent agreement has optically-thin approximation for radiation heat
been found between the temperature distributions transfer.
based on the nongray wide-band model and experi- Analysis reveals eight dimensionless parameters
mental data (Fig. 24). Examination of Fig. 24 reveals which control the system under investigation. Five
that the wide-band model yielded results that were parameters, i.e. the mass consumption number, r; the
superior to those that excluded radiation-convection mass transfer number, B; the Prandtl number, Pr; a
interaction. It is evident from the figure that dimensionless heat of combustion, D o the fuel
radiation-convection interaction lowers the pre- surface temperature, 0,., are the combustion groups,
dicted temperatures in the high-temperature region and the three radiation groups, the conduction/
of the boundary layer near the flame front and raises gaseous radiation parameter, N~, the conduction/
the temperature in the cooler region near the edge of ambient radiation parameter, N 2 , and the fuel
the boundary layer. Furthermore, this interaction surface emissivity, e,., are required to describe the
effect increases for larger fuel flow rates. 246 The effect combusting-radiating system. The parameters D c
of radiation interaction can be interpreted to result and 0w, which were of secondary importance in
from a transfer of energy by gaseous radiation heat nonradiating systems, emerge from the analysis with
transfer from the hotter region to the cooler portion new significance, dominating the parameters N~, N 2
of the boundary layer, thus reducing the higher and ew.
temperatures and raising the lower ones. The effect of radiation on the pyrolysis rate and
An attempt was made to determine an arbitrary unburned fraction of total pyrolyzate is shown in
value of the Planck mean absorption coefficient ~ Fig. 25 for axisymmetric combustion. 248 The pyroly-
which when used in a gray-gas model would yield sis rate is seen to increase strongly with increasing
temperature profiles matching the experimental dimensionless heat of combustion, D c. This is
results. 246 Although the results of Fig. 24 demon- primarily because of an increase in flame temper-
strate that matching values of ~e may indeed exist, a ature (01 ~ O,.Dc). Like the mass fraction of fuel at the
different value of a mean absorption coefficient was surface, the dimensionless flame temperature 0y,
required for each fuel-flow condition. It has been which for non-radiating flows can be determined a
shown in Section 4.5 that by changing the mean priori from measurable quantities, depends on all
138 R. VISKANTAand M. P. MENGOq

I.o .... i.o

Rodiotin9
------ Non.r(idiolin9 t

o.s -~
o.s N

0 0
I.O IO
DIMENSIONLESS HEATOFCOMBUSTION,Oc
FIG. 25. Pyrolysis rate and unburned pyrolyzate vs dimensionless heat of combustion for axisymmetric
flow with B= 1.0, r=0.22, 0,,.=2.0, N~ = 0.05, N 2 = 5 0 . 0 , and e = 1.0. 248

eight parameters and is not predetermined. The butions of soot volume fractions and CO 2 and H 2 0
pyrolysis rate with radiation is lower due to the net concentrations. The utility of the analysis will then
efflux of radiation at the surface. In general, the come from both the proper quantification of radi-
influx of gaseous radiation is insufficient to cancel ative effects in opposed-flow diffusion flame experi-
the efflux of surface emission, hence a lower pyrolysis ments and from the use of such systems to refine
rate results in comparison to non radiative combus- techniques for incorporating radiation in combus-
tion. Lower pyrolysis rates may result even when a tion modeling.
net influx of radiation prevails because of the
decrease in conduction caused by the lower flame
5.6. EJJect oJ Radiation on a Planar, Two-Dimensional
temperature due to radiant loss from the combustion
Turbulent-Jet DiJJusion Flame
zone. At low Dc, the reaction releases little energy to
counter surface emission losses, giving low pyrolysis A simple combustion situation has been modeled
rates, whereas at large Dc much energy is released to assess the importance of thermal radiation in
which easily overcomes surface losses and yields establishing temperature distribution in a turbulent
large pyrolysis rates. 2as The net effect of radiation on diffusion flame. 25 Although, turbulent diffusion
pyrolysis appears to be low for several reasons. The flames have been extensively studied by Bilger and
properties of real opposed diffusion flames are not his co-workers) 51- 253 they have not considered the
yet sufficiently well known to give accurate para- effects of radiation. However, radiation heat transfer
meter values. Those chosen for the calculations (see modifies the temperature distribution which, in turn,
caption of Fig. 25 and others in Ref. 248) may not be affects the combustion process. Small changes in
sufficiently realistic as they all tend to underestimate peak temperatures have a large influence upon nitric
the differences between non-radiating and radiating oxide production for a given residence time. It is of
systems. interest to determine how the various control
As data on radiative properties of stagnation- strategies such as lowering combustion air preheat or
point flames become available, the approximation of recirculating exhaust products into the combustion
a constant absorption coefficient should be replaced air affect the unwanted nitric oxide emissions and
with a nonuniform one based on measured distri- the desired radiation heat transfer.

BLACKPLANEWALL
!11111111111111111111111111111/6
AIR t
i JET
FUEL ~f MIDPLANE

Am ///7"//////////////////////// ///
BLACK PLANEWALL
FlG. 26. Schematicaldiagram ofaphme, radiatingjetconlinedbetweentwoparallelplates.
Radiation heat transfer 139

The physical model of the problem analyzed by G4


James and Edwards 25 is shown schematically in
,,, .t .,o ..o
Fig. 26. A planar jet of methane is injected with .... . s A . peso. H ,1T/'~
velocity ut,~,l into a stream of air flowing with -.- s~e,. ,~o. I ] :W ',.~
velocity u,~r parallel to the fuel. Diffusion-controlled I
!
o.2 ,..,ooo iYl "\
combustion occurs in the mixing region of the jet. "s "L ,1 v ",~
Plane-parallel, isothermal and black walls sym-
metrically located above and below the jet form the _sJ ~.J.J I%~J I "'-'~- . . . .
~ 0.0
combustion chamber. A soot-free flame is assumed to
I I
exist so that molecular gas bands determine the
thermal radiative transfer to the walls. Boundary- _ x/O - 50
,z, ~4i
layer approximations were used to simplify the - ~1~\ -
conservation equations, and nongray radiation
described by the exponential wide band model for ,q
molecular gas band radiation was added to the
energy equation. The model equations for turbulent
combustion of methan in a planar, enclosed jet- - -

diffusion flame were solved numerically.


The analysis demonstrates that realistic nongray 0.0
1.0 2.0 3.0 4~
radiative transfer calculations can be coupled to an
implicit numerical method for solution of the highly WAVELENGTH (pm)
nonlinear partial differential conservation equations FIG. 27. Spectral radiation intensities ~ r radial paths
without undue expenditure of computation time. The through aturbulenthydrogen:airdiffusion flame. 25s
results of computations have shown that the larger
channels (A= 1 m and 10 m) have markedly lower
peak temperatures because of greater gaseous radi- Estimates of spectral intensities emerging from
ative transfer. It was also found that a given flames, based on predicted mean scalar properties,
reduction in minimum combustion temperature to are typically within 20-30 % of the measurements of
reduce nitric oxide formation could be accomplished well-defined laboratory flames. 31,25'*.255 This is com-
with a much less detrimental reduction of heat parable to the uncertainties in the narrow-band and
transfer by recirculating exhaust product into the flame-structure models. Measured and predicted
combustion air than by reducing preheat. spectral radiation intensities for a turbulent hydrogen/
air diffusion flame are given in Fig. 27. Results are
5.7. Radiation Ji'om Flames shown for horizontal radial paths through the flame
at x/D=50 and 90, the latter position being just
Gas- and liquid-fueled flames have numerous below the flame tip. Predictions use both time-
applications and flame radiation is an important averaged scalar properties along the path and
aspect of heat transfer in furnaces, internal combus- stochastic methods which take into account turb-
tion engines, aircraft propulsion systems, flares, ulence/radiation interactions. The stochastic method
unwanted fires, etc. This has motivated many studies models the interactions by assuming that the flow
of flame radiation and comprehensive, up-to-date field consists of many eddies which are uniform and
reviews are available.'*a'239.25'*'255 The issues of statistically independent of each other. Eddy length
concern here are nonluminous and luminous radi- varies along the path length, and time-averaged
ation from flames, prediction of radiation character- probability density (PDF) of mixture fraction f for
istics given the instantaneous scalar structure, and each eddy is randomly sampled and scalar properties
turbulence/radiation interactions in simple laboratory are found from the state relationships at the
flames. corresponding value of J~ Once the scalar properties
Significant progress has been made concerning are known, the RTE is solved. The details of solution
structure and prediction of radiation intensity of can be found elsewhere. 31 Spectral radiation inten-
nonluminous flames. Narrow band-model predic- sities (Fig. 27) are dominated by the 1.38, 1.87 and
tions T M for nonisothermal mixtures of CO2, H 2 0 2.7/~m water vapor bands in the range of 1-4/~m
and CO are in good agreement with the measure- shown. The stochastic method yields spectral inten-
ments. The total transmittance nonhomogeneous sities which sometimes are about a factor of two
model (TTNH) of Grosshandler 256.257 has been higher than the mean property method with the
found to be about 500 times faster than narrow-band measurement generally falling between the two
models. The model has been applied to several predictions. These results suggest significant effects
realistic combustion examples containing variable of turbulence/radiation interactions. Findings for
concentrations of CO2, H20, CH4, CO and soot. It carbon monoxide/air and methane/air flames, how-
was found to be usually within 10% of the more ever, show smaller effects for turbulence/radiation
accurate computation. 2s 7 interactions. 254.255
140 R. VISKANTAand M. P. M~,~GO~

Work on luminous flames has been limited. results have been obtained for other laboratory
Similar results to those presented in Fig. 27 have diffusion flames. Excellent agreement has been ob-
been reported by Gore and Faeth (cited in Refs 254 tained between measured and predicted radiative
and 255) for a turbulent ethylene/air diffusion flame. heat flux distributiotrs parallel to the axi's of
The spectra are dominated by continuum radiation turbulent carbon monoxide/air diffusion flames. 259
from soot, however, the effects of 1.38, 1.87 and The mean property predictions agree very well with
2.7/~m gas bands of the H 2 0 and the 2.7 and 4.3 #m the measurements because the effects of turbulence
gas bands of CO2 can still be seen. In this case the radiation interaction are small. The analysis correctly
mean-property method has provided the best quan- predicts maximum heat fluxes near the flame tip as
titative agreement with the data, but the agreement is well as the effects of burner flow rate.
considered to be fortuitous in view of poorer Discussion of the effects of turbulence/radiation
extinction predictions obtained using the ap- interactions has been given by Faeth et aL 31'255 The
proach. 255 The predictions of continuum radiation available results show that the interactions are very
are very sensitive to local temperature estimates, and significant for hydrogen/air diffusion flames, with
the assumption optically-thin radiative heat losses stochastic predictions being as much as twice the
are quite crude. Differences between mean property mean property predictions. 2sa In contrast, turbulence/
and stochastic predictions suggest significant effects radiation interactions caused less than a 30~o
of turbulence/radiation interactions in luminous increase in spectral radiation intensities for carbon
flames. More exact coupled structure and radiation monoxide/air and methane/air diffusion flames. This
analysis could modify the relative performance of the difference is attributed to the relatively rapid vari-
mean-property and stochastic methods and suggest ation of radiation parameters (water vapor concen-
that presently available models must be improved. tration and temperature) near stoichiometric con-
Measurements and predictions of total radiative ditions for hydrogen/air diffusion flames. The
heat fluxes to points surrounding the turbulent stochastic methods at.254~255 have many ad hoc
hydrogen/air, 258 carbon monoxide/air, 259 methane/ features and additional fundamental research effort
air, 26 and ethylene/air T M diffusion flames have been is needed to develop more reliable methods not only
made. The discrepancies between the measured and for small laboratory flames but also for scaling large
predicted total radiation heat fluxes along the axis of flames containing soot.
a turbulent methane/air diffusion flame (Fig. 28) are
within the order of 10-30 %. Such levels of error are
similar to the differences between prediction and 5.8. Combustion and Radiation Heat TransJer in a
measurement for the spectral intensities. Comparable Porous Medium

Although flame radiation plays an important role


in combustion systems, a furnace requires a sufficient
1.0 I I I I volume for the heating chamber to increase the
opacity of the flame and furnace for effective
radiation heat transfer to the load. Moreover, the
load must be placed away from the reaction zone to
0.8 prevent the emission of unburnt species when its
surface temperature is low. These factors make it
difficult to reduce the size of a combustion chamber
E
appreciably. Echigo 262,26a has shown that a porous
0.6 medium of an appropriate optical thickness placed in
a duct is very effective in converting enthaipy of a
X flowing gas stream to thermal radiation directed
:3
.-I
t,I. toward the higher temperature side. Successful appli-
0.4 cations to an industrial furnace, 262'263 to a combus-
m.m
tor of low calorific gas, 264 to a water tube by
W combustion gases in porous media, 265 and to other
V- systems 26a have been reported. The thermal structure
<r, O.2 in the porous medium with internal heat generation
due to chemical reactions has been studied analytic-
ally and experimentally,265'2~6 and a review of the
work is available. 26a'26~
O.0t- A one-dimensional model in which radiative
0 800 1600 2400
transfer in a gas-solid two-phase system is treated
AXIAL DISTANCE (ram) rigorously has been constructed, and extensive
FIG. 28. Total radiative heat flux distribution along the axis numerical calculations have been performed for a
of turbulent methane air diffusion flames at NTP. -'' radiation controlled flame. 267,26a The combustion
Radiation heat transfer 141

I f I I I I Significant energy recovery has been achieved from


1200 the burned gas to preheat the combustible mixture
prior to entering the reaction zone by propagation of
thermal radiation against the flow direction. This is
deafly shown in Fig. 29 which compares the
800 predicted and measured particulate-phase (Te) temp-
eratures in the system. 26s In the figure, both the
distance x and also the optical depth z along the
4O0 . . . . 1538 W combustion system are used as the abscissa. The
o .... 1025 W
"~ ~ ' - - 513W results show that as the combustion load increases,
both the measured and calculated temperatures
increase uniformly. This is a consequence not only of
O I I I I I I
-IO 0 1(3 20 30 40 50 the relative reduction of heat loss in comparison to
I x {mm} heat generation during combustion but also due to
~I P.-t aM-= I~ PM-n the essential nature of radiation heat transfer.
-6o , 3 5 ,,.35
r

FIG. 29. Comparison of measured and predicted temper- 6. APPLICATIONS TO COMBUSTION SYSTEMS
ature structures in porous media for different combustion
loads. 2~8 The lower abscissa scale r is optical depth based The advent of more powerful digital computers
on the ~bsorption coefficient of the porous medium, and has provided the means whereby mathematical
PM-I, PM-I1 and PM-III are the abbreviations for porous
media I, I! ~md II!, respectively. modeling can be applied to combustion system
problems to facilitate the arduous task of their
design. This is now of great interest in view of the
mixture flows through a porous medium and the current demands which system designers are required
combustion reactions take place in the medium. The to meet--in particular, efficiency of combustion at a
results of comprehensive calculations show that the wide range of operating conditions and strict control
thermal structure (profiles of temperature, local of pollutant emissions. The latter has become
radiation flux, etc.) in the high porosity medium increasingly stringent in recent years for economic
depends strongly on the absorption coefficient and and political reasons. The present trend is away from
total optical thickness of the medium as well as the the traditional cut-and-try methods, which are ex-
position of the reaction zone. G o o d agreement pensive and do not necessarily produce the optimum
between predicted and measured temperature distri- design, toward fundamental modeling of the physical
butions has been obtained and a drastic temperature and chemical processes occurring within the combus-
decrease in the porous medium has been re- tion systems. Multidimensional modeling of two-
vealed. 266"26s The results have also revealed remark- phase combustion is being approached with the aim
able heat transfer and combustion augmentation. of producing algorithms based on fundamental

I Two Phase Fluid


Mechanics
(Turbulent)
Phase
Chemical Transitions
Kinetics (Evaporation)
(Condensation)

Gas
Nucleation Particle
Interaction

Par title Phase


Reaction Gas Phase
Devolatilizofion) Reaction
eter. Oxidation)

Heat Transfer
(Convective)
(Radiative)

FIG. 30. Schematic representation of submodels for combustion of coal.


142 R. VISKANTAand M. P. MENGi3t~

principles which can correlate all of the details of radiation model, through the radiative properties of
combustion systems. 3,269-272 The predictive pro- combustion products, to the mathematical transport
cedures for a combustion system model require model to predict the temperature and radiating
theoretical and empirical inputs to describe turbulent species concentration distributions. With the pres-
flow, chemical kinetics, thermodynamic and thermo- ently available algorithms, 3"5'269-271 the latter type
physical properties and other transport processes, problems require an iterative solution procedure
including radiation heat transfer (see Fig. 30). which is rather time-consuming.
This section of the article discusses application of A validated computer model has been used to
the methodology described in the previous sections construct a detailed energy flow (Sankey) diagram for
to practical combustion systems. The emphasis is on an industrial furnace. 273 The diagram (Fig. 31) shows
the methodology and radiation heat transfer results that more than half of the heat to the load comes
rather than the application of mathematical tech- from the refractory wall. Of the balance, part is
niques for design and performance calculations of convection (4~), part is direct radiation from the
practical systems. Even with the advances in main- flame (6 ~), and part is flame/wall radiation absorbed
frame computers the difficulty of treating infrared by the gas which has been re-radiated by the wall
radiation transfer rigorously in nonhomogeneous (6 9/o). The furnace shows a thermal efficiency of 35 ~o
gases containing particles lies primarily in the with the typical high flue loss and indicates the
enormous complications introduced by selective importance of the wall-to-wall re-radiation effect.
gaseous emission and absorption of radiation as well With the exception of different magnitudes, Fig. 31
as scattering by irregular-shaped particles. Because of shows a typical pattern for all industrial natural gas
this complexity practical simplifications are necess- and oil fired furnaces. As the flame becomes more
ary to keep the calculations at a reasonable level. As opaque and/or the wall temperatures drop there will
a compromise between desired accuracy and compu- be obviously more radiation from the flame and less
tational effort, practical methods which are also from the wall. Also, as the wall temperatures drop
compatible with the numerical algorithms for solving there will be smaller radiation exchange between the
the transport equations are stressed, and radiation walls and the load.
heat transfer in several different combustion systems The close examination of Fig. 31 clearly indicates
is discussed. The body of literature concerned with why there has been so little attention given to the
modeling and evaluation of combstion systems is calculation of convective heat transfer inside fur-
very large, and it is not practical in an article of naces. In industrial furnaces convective heat transfer
limited scope to discuss even the more recent works. usually accounts for a very small fraction of the total
Most of the work reported has stressed modeling and heat transfer to the load. Local convective heat
evaluation of chemically reacting turbulent flows and transfer coefficients have been measured at a surface
combustion and much less radiation heat transfer. heated by gases 274 and empirical correlations for the
The emphasis in this review is on the latter. average Nusselt number have been reported for
differently-directed gas streams incident on the
load. 274-276 An interesting finding of the experi-
6.1. Industrial Furnaces mental study 274 was that in the absence of combus-
One of the important parameters in assessing the tion the average heat transfer coefficient at the load
performance of an industrial furnace is the heat flux surface was about 35 W/m2K, while in the presence
distribution to its thermal load (sink). Methods of combustion the values were from 80 to 120 W/m2K,
based on fundamental principles are now available suggesting almost a threefold enhancement of con-
using numerical techniques and digital computers, vective heat transfer by combustion.
that permit determinations to be made for both gas- Radiation in furnaces predominates over con-
and oil-fired industrial furnaces. In such furnaces vection; therefore, more emphasis has been given to
heat transfer to the load is predominantly by thermal radiation over the years and the radiative transfer
radiation. The problems associated with prediction theory has been much more fully developed,4-5 2 7 7 - 2 8 1
of radiation heat transfer within the combustion and presently capability exists to predict simultaneous
chamber can be divided in two main types: three-dimensional flow, heat transfer and reaction
rates inside furnaces. 269'27! However, the theory has
(a) Evaluation of radiation heat transfer at all outstripped experimental validation, which is in a
locations in the enclosure if the temperature much more primitive state, but even in this area a
distribution and radiative properties of the number of papers describing direct comparisons
combustion products are known; and between predictions and experimental data have
(b) Evaluation of radiation heat transfer as well as appeared, t 6 9 , 2 8 2 - 2 8 8
temperature and radiating species concentration The results obtained for a model furnace using
distributions. the phenomenological furnace-performance equations
Problems of type (a) are more straight-forward and have been used to determine the relative importance
require development of radiation heat transfer models~ of the model parameters. 2s Analysis of the results
Problems of type (b) require the coupling of the led to the conclusion that the flame emissivity was of
Radiation heat transfer 143

~/AuT//~~~~ CONVECTION TOI.OAD4/mL

FItJ. 31. Energy flow (Sankey) diagram for one operating point of un industrial furnace, illustrating the
four different contributions to output .'rod the effect of wall-to-wall radi~,tion exchange, z-3

Type 1: Stirred Vessel


rather specific conditions; and some aspects of the
analysis could be considered arguable.
'' ,--t- The models for analyzing heat transfer in
industrial furnaces are of three types (see Fig. 32):
(1) the "stirred vessel" (zero-dimensional)
mode145.273.277.278,281,285-289 which yields only the
~Heot Flux Oistri~ total heat transfer rate without providing infor-
Type 2: Plug Flow mation on the local heat flux distribution, (2) the
"plug-flow" (one-dimensional) model z76,2s,zs2-2ss
which is capable of predicting the local heat flux in
the furnace along the flow direction, and (3) the
multi-dimensional model 269.271 which can predict
two-dimensional heat flux distribution at the load
surface. The first two models are being used routinely
in engineering design calculations, and these models
Type 3: Two - Dimensional are discussed here in greater detail.

6.1.1. Stirred vessel model


',x~- .~:- F-~T-j---:t:_F-_a..i Let us consider a schematic diagram of a furnace
(Fig. 33) and apply the "stirred vessel" model to
calculate the heat transfer rate to the load. According
to the model '.5.277,278.2al the combustion products
Heo! Flux Oistrib. are assumed to be gray and at a uniform temperature.
HG. 32. Schematic representation of it furnace illustrating The temperature and the radiative properties of the
different heat transfer models.2as load and of the refractory walls are assumed to be
uniform but different. A steady-state, overall energy
second-order importance. The other factors (in the balance on the load can be written as
order of decreased importance): heat transfer to the
load (sink), excess air, process temperature, flame/load /~/1 -/~/2 = Q, + Qe (6.1)
temperature difference, load absorptivity and wall
losses, were of greater significance. These conclusions Here/~/~ and/:/2 are the nthalpy inflow and outflow
were reached, however, by generalizing from some rates. Within the framework of the zonal approxi-
144 R. VISKANTAand M. P. ME~Gi~t~

,0,, Qs
F s --
Y,,'/,// ,/////'///////// A~_,,a T~
Waste Gases
~aml~stloa ==~.
~,.r=.~ = [ 0 ~ - O] + ( S t ' K o X O m - 0,)]. (6.5)
Products
o., T,
Fuel fit AIr For the special case when convective heat transfer to
/~> ~o. e/r,,A ' the load is negligible in comparison to radiation
//I Lo,, (St=0), Eq. (6.4) simplifies to
/
7111111111111111~
Ko(1 - 0m)= 0~-- 0]. (6.6)
FIG. 33. Schematic diagram of a stirred furnace model.
By eliminating the mean combustion-product tem-
perature, the dimensionless heat transfer rate can be
expressed as
mation for radiation heat exchange, the heat transfer
rate to the load can be expressed as 277'2al
r ~ = Ko[1 - ( F, + 0 ~)z/'*]. (6.7)

O~---A, [h(T~ - T~)+ ,fr _ ,,a(T~ - T])] (6.2) Extensive calculations have been reported for the
dimensionless mean gas temperature and heat trans-
where h is the average convective heat transfer fer rate and the results can be found in the
coefficient at the load, and " ~ s - m is Hottel's radiation literature. 2aS-za7 Experiments have also been per-
exchange factor or A ~ - ~ _ m is the total radiation formed and compared with model predictions. 2s6'2a7
exchange area. This factor is a rather complicated
Figure 34 shows a comparison between the measured
function of the gas emissivity, wall emissivity and the and the calculated average heat fluxes in an experi-
sink-to-refractory area ratio, and expressions are mental combustion chamber having a 1.25 m long
available in the literature. 45'287 The heat losses
firing space and two different cross-sections (0.4 m x
through the walls of the furnace can be expressed as 0.4 m and 0.4 m x 0.8 m). The results show that the
stirred-vessel heat transfer model can be successfully
Qt = UoAo(Tm - T,) (6.3) applied to those furnaces in which there is no
appreciable axial drop of the mean gas temperature.
where U0 and Ao are the overall heat transfer This condition is roughly met in combustion cham-
coefficient and area of the refractory walls, respec- bers fired with high-velocity burners and in furnaces
tively; and T,, and T, are the mean combustion where the flame length is approximately equal to the
products and ambient air temperatures, respectively. furnace length. Under these conditions, a maximum
Substitution of Eqs (6.2) and (6.3) into Eq. (6.1), and error of _+20% can be expected in calculating the
assumption of negligible wall heat losses allows the absorbed heat flow to the load being heated. In
resultant equation to be written as predicting the energy consumption of the furnace,
this would mean a maximum error of _+10 ~oo.286
+- 0~-0,,, = ( 1 / K o ) ( 0 ~ - 0 ~ ) (6.4)
l + St 1~ l + St
100 F~naee Crog-Seellon In mmZ=LOO,t.O0 ~Or,9~
Io, 06 V
where the dimensionless variables and parameters
are defined as

0 =T" 0 T~ ;nCpm
60
" 7,; A ~ . _ ma T 3 -.. \

hA~ "1- "~ d" k


St= :~.
rnC pra

In this equation, T~ is a fictitious gas inlet tempera-


ture in which the heat losses through the walls of the
furnace have been accounted for; m and q,,. are the 0
gas mass flow rate and the mean specific heat of the 0 200 too 600 oCo iOoo 1200 'v,OO
gas, respectively, and K o and St are the Konakov and Surface Temperature (~)
Stanton numbers, respectively. The heat transfer rate FIG. 34. Comparison of measured and predicted average
to the sink, Eq. (6.2), can be expressed in dimension- heat fluxes in a furnace as a function of the load
less form as temperature, z8~
Radiation heat transfer 145

6.1.2. Plug flow model where

A schematic diagram of the "plug-flow" model is


x T~ 0 _ 7 " ;
shown in Fig. 35. The temperature of the gas "= '
(combustion products) is assumed to depend on the
coordinate x in the flow direction. This means that
the plug flow model can be considered to consist of Ko= DiCpm h WL
St . . . . .
an infinite number of stirred vessels. The temperature WLi,.fl T 3 ~ncp~
and the radiative properties of the load and walls are
assumed constant but different. Based on a gray-gas UoPoL
and zonal approximation for radiation heat exchange, dPo -
thcmn
the steady-state energy balance on a control volume
of gas of length dx gives Analytical solutions of Eq. (6.9) and its special forms
have been obtained and graphical results re-
ported.29'291
dT~(x) . , ,
thepm dx - - W{e,a[T,(x)-- T,] Extensive numerical calculations of the gas tem-
peratures along the furnace using the stirred-vessel,
+h[T~j(x)- T~] } - P o U o [ T ~ ( x ) - T j (6.8) stirred-vessel-cascade, plug flow and the modified
zonal models have been reported for furnaces having
constant and varying sink temperatures. 292'293 A
where W and P0 are the furnace width and perimeter, comparison of temperature distributions using five
respectively, and lg is the effective gas emissivity zones (sections) along the furnace is given in Fig. 36.
which accounts for the refractory walls and other The published results show that as the number of
surfaces in the furnace. In dimensionless form, the sections in the furnace increases, the temperature
energy equation for the gas temperature can be distribution predicted employing the stirred-vessel-
written as cascade and the modified zonal models approaches
the temperature calculated using the plug-flow model,
As expected for a single section along the furnace, the
stirred-vessel and the modified zonal models predict
dO._ (1/KoXO4_O~)_St(O_O~) practically identical gas temperatures in the furnace.
d

-- ~0(0.--0~) (6.9)
6.1.3. Multi-dimensional models

~ dQ,,
The computational methods which have been
developed are able to complement, but not replace,
empirically based design procedures. This is because
H Products IldO~ H U I'T! chemically reacting turbulent flows are not fully
understood, and it is proving particularly difficult
to eliminate the deficiencies of existing turbulence
.... ........ -.:'H models. In the absence of reliable turbulence models
it is hardly possible to subject any of the ever-
FIG. 35. Schematic diagram of a plug flow model. increasing number of combustion and radiative

1800
\\ ' ' {o)' '
i i i

\
,~. Plug Flow %--Plug Flow
1600
\
\
ca 2x_~~ .,. Stirred Vessel
1400 , S f i t r e 1 % % ;!.,... / :aecade
Vessel ~ ~.
E \
1200 Zonol ~ Stirred Vessel

O Zonal
it\x)
o i
02
014 i
0.6
i
08 0
1
0.2
i
0.4 0.6
i
0.8 LO
x~
FIG. 36. Comparison of gas temperature distributions along a one-zone la) and five-zone (b) furnace
predicted by different models: Ko= 1, K=0.1 m- ] i:~j=0.104,~==0.8, h/h=2/l, l/h=20/l, T~= 773 K. 2<j3
J?gCS 13:2-D
146 R. VISKANTAand M. P. MENG(~(;

transfer model proposals to a stringent assessment. 250 i , i , , i ,

Nevertheless, two-dimensionaP ,294.295 (among others) ------ / T I i " I000 K


and three-dimensional 269.271.296 combustion valid- 2O0
ation studies reveal, for gaseous combustion at least,
that predictions which are obtainable are sufficiently 150

reliable to be of interest to combustion engineers.


General computer-based procedures for the predic-
tion of gaseous-fired rectangular and cylindrical

.'Wilh
5O
combustion chambers have been developed and a
review is available: The zonal, flux, discrete- 0 Tuth/RQa. Inlw.
ordinates and first-order spherical harmonics ( P : Willloul Turb./Ro4. I n t l r .
approximation) methods have been assessed. For -50 I I I I I I I I I

natural gas and oil fired furnaces only three species 0 0.2 04. 0.6 0.8 1.0

(CO2, HzO and soot) contribute significantly to the x/L


transport of radiation in the infrared. The compu- FIG. 37. Effect of preheated air fuel mixture temperature
tations have been carried out on the gray or at most and turbulence, radiation interaction on heat flux distri-
on a weighed sum-of-the-gray gas bases. Reasonable bution along a two-dimensional furnace burning methane:
H= 1 m. L=5 m, ~ = 1500 K. e~=0.8, e, =0.6. 3
agreements are reported between measured and
predicted fluxes (see Ref. 5 for comparisons). Unfor-
tunately, the original references includes little detail
on how the mean absorption coefficients needed in equations constituting subsidiary models of turbu-
the radiative transfer models have been determined. lence, chemical reaction and radiation heat transfer
It is suspected that the authors had to do consider- phenomena. A six-flux, gray gas model is used to
able "fine-tuning" of these model parameters to bring predict radiative transfer. Computer-memory limit-
about good agreement between model predictions ations restrict the amount of geometrical detail that
and data. The sensitivity of the results to radiative can be included and prevent the use of a finite-
properties have already been discussed in Section 4.5. difference grid having the desired fineness. The model
It should be pointed out that in the studies is validated against experimental data acquired on
discussed by Khalil 5 and others 269 the emphasis has two large, natural gas-fired furnaces.
been on modeling chemically reacting turbulent flow Recently, the effect of turbulence/radiation inter-
and combustion and much less on realistic modeling action in a two-dimensional, natural gas-fired, indus-
of radiation heat transfer. The general prediction trial furnace has been examined. 3 Based on an
procedures which describe the computation of flow, approximate analysis of radiative transfer, the results
reaction, and heat transfer in the combustion region of calculations show that the effect of turbulence/
of a typical, natural gas-fired industrial glass pro- radiation interaction on combustion and scalar
ducing furnace are sufficiently developed to consti- properties is small for a preheated fuel-air mixture
tute a useful design tool. 269 Economic handling of when the flame occupies a small volume of the
three-dimensional geometric features is considerably furnace. However, when the flame occupies a large
enhanced by the use of special grids and the separate volume fraction of the combustion chamber the
calculation of the burner and bulk combustion interaction is quite significant. Another reason why
chamber regions in a manner which takes into the effect of the interaction is larger for T;= 300 K
account the differing features of their flows. The than for T~=1000 K is because the temperature
predictions demonstrate the value of computations fluctuations are larger for the former case. The effect
to furnace designers for the range of operating of the interaction on the total heat flux along the
parameters. Recently, the radiative transfer has been furnace shown in Fig. 37 clearly indicates the need to
treated in sufficient detail using the discrete transfer account for turbulence when predicting radiation
method which contains some features of the zone, heat transfer in large, high-temperature combustion
discrete ordinates and Monte Carlo procedures. 24 systems. The net local heat flux to the sink (load) can
The combustion products are treated as gray and become negative for the case when the turbulence/
scattering by particles, such as soot agglomerates has radiation interaction is neglected, because the as-
been neglected. sumed sink temperature (Ts= 1500 K) is higher than
A detailed discussion of analytical modeling of the local effective temperature of the combustion
practical combustion chambers and furnaces, in- products.
cluding a very extensive review of the literature has
recently been given by Robinson. 271 A three-
6.2. Coal-Fired Furnaces
dimensional mathematical model is constructed of a
large tangentially-fired furnace of the type used in Radiation heat transfer in coal-fired furnaces has
power-station boilers. The model is based on a set of received considerable attention for more than 60 yr
13 differential equations governing the transport of because of the realization that it is the dominant
mass,momentum and energy, together with additional mode of heat transfer in such systems. The earlier
Radiation heat transfer 147

work on the subject has been discussed by Doleza1297 matter deposited onto surfaces of coal-fired furnaces
and more recent studies have been reviewed by can greatly affect radiation heat transfer due to the
Blokh. 4 The latter volume in particular contains a alteration of its emissivity. 7s Mineral matter and ash
large body of fundamental radiation property data, deposited on walls of the tubes can also increase
measured spectral and total incident radiation fluxes greatly the thermal resistance to heat conduction
along the height of different capacity furnaces as well across the deposit, and some simple conductance
as empirical correlations for analyzing the thermal models have been developed.'*
performance of coal-fired boilers. An up-to-date Data for soot, carbon and coal refractive indices
discussion of coal combustion models in which are generally (but not necessarily very accurately)
radiation heat transfer has also been considered is available,'*'64 but significant uncertainty exists in the
available. 3 Despite the considerable progress in the particle concentration and size distributions. In
development of analytical methods of engineering gasifiers and staged combustion systems, which
science and despite an increasing understanding of operate fuel-rich for nitrogen oxide pollutant control,
fundamental combustion processes, the design or soot radiation may be particularly important. Unless
performance predictions of coal-fired furnaces may the soot-volume-fraction distribution in the medium
still be considered as an art based primarily on is known accurately, radiation heat transfer to the
empirical knowledge and the ingenuity of the com- chamber walls can not be predicted with confidence.
bustion engineer. This is particularly true for large Fly-ash particles greatly influence the radiative
boiler furnaces because of their extremely compli- properties of the flame and of the combustion
cated geometry and boundary conditions 4'272'29s as products in a pulverized-coal fired furnace. Data for
well as the lack of confidence in the existing fly-ash are much less certain. 4'79-83 There is signifi-
analytical methods. Scale-up and advanced perform- cant variation in the refractive indices of pulverized-
ance analyses of boiler combustion chambers have coal and fly-ash with the type of coal, mineral matter
been developed272 using laboratory and/or small in the coal, as well as the combustion process itself.
model furnace data. In spite of major improvements Experiments have revealed that the refractive index
in the analytical methods for predicting the perform- of fly-ash particles formed during the combustion of
ance of coal-fired furnaces 3'272 there is still distrust even one coal shows quite large differencesfl 7 Lowe
by practical furnace designers of the analytical et al. 3 have shown that in large boilers fly-ash
methods because of geometrical restrictions, problems exerts a much greater effect on heat transfer to the
of stability, complexity of the new methods, limited heat-absorbing surfaces in a furnace than the aero-
applicability of the models, etc. dynamics and kinetic characteristics of a pulverized
In this section we discuss the use of more recent coal burn-out. Radiation from fly-ash particles
models to predict radiation heat transfer in relatively exceeds substantially the contribution of both tri-
simple furnaces, for the purpose of gaining improved atomic combustion gases, as well as char and soot
understanding of radiative transfer and of the particles'* Contribution to radiative transfer by char
relative importance of the model parameters. It is particles is essentially over the length of the flame. At
hoped that this would provide the bridge between the the end of the furnace the concentration of the char
scientific community which is developing compre- particles is small, and there they exert very little effect
hensive combustion system models and furnace on the radiation heat flux at the wall.
designers who are attempting to solve practical
problems based on empirical knowledge. Reference is
made to literature which discusses methods for 400
evaluation of thermal performance of large boiler g
furnaces.
Detailed reviews of radiation heat transfer in
300
pulverized coal-fired furnaces are available.4"272"299 c+s+g
Radiation heat transfer in furnaces is due to gaseous
and particulate contributions. Emissivity data for the
qr, z,,,
(kW/m 200
major emitting gaseous species CO2 and H 2 0 are
generally adequate. 4.64 Other gaseous species (e.g.
CO, SO2, NO, N 2 0 ) are usually of secondary
importance because of low concentration. Local I00
variations in gas temperature and species composition
are subject to more uncertainty than the emissivity
data. Contributions to particle radiation in pulver- 0
0 2 4 6 8 I0
ized coal-fired systems usually results from coal z On)
(char), soot andfly-ash. Information required for
FIG, 38. Effect of combustion products composition on the
predicting radiative transfer includes different particle radiation heat flux distribution along the wall of a
concentrations, size distributions, complex indices of pulverized coal-fired furnace; (c=coal, f=fly-ash, s=soot,
refraction and temperature. 2~'* Finally, the mineral g= combustion gases), for soot J,, = 2 m - 1.2,4
148 R. VISKANTAand M. P. MENGOt;

Radiation heat transfer in a cylindrical, pulverized Current reviews of coal-fired combustion models
coal-fired combustion furnace has been predicted are available 2'3 there is no need to repeat these
both on a gray 214 and nongray s9 basis. The comprehensive discussions. Recent radiative transfer
calculations were carried out by assuming the modeling for inclusion in comprehensive multi-
temperature and radiating species concentration dimensional combustion codes has focused on more
distributions in the furnace. The radiative character- efficient differential and flux methods, 31-303 but
istics of the coal particles were predicted from the there are exceptions. For example, Truelove 34 used
Mie theory, after first assuming a coal particle size a discrete-ordinates method, which is more time-
distribution. Details of radiation heat transfer and consuming to evaluate; however, to simplify the
sensitivity calculations can be found elsewhere. 214 procedure the gas was considered to be gray and the
The contributions of the different constituents (coal, particles were assumed to be black and nonscat-
fly-ash, soot and combustion gases) on the local tering. The classical Hottel zonal method is compu-
radiative flux along the furnace are shown in Fig. 38. tationaUy inefficient for use in multidimensional
It is clear from the figure that neglect of the fly-ash codes. In addition, there are conceptual and numer-
contribution and inclusion of soot absorption yields ical difficulties in adopting the method when aniso-
a dramatic change in the radiative transfer in the tropically scattering particles are present in the
medium and at the cylindrical walls (see curves combustion products.
denoted as c + f + g and c + s + g ) . The main reason Available computer models for scale-up and per-
for this discrepancy is the replacement of strongly formance predictions of boiler combustion chambers
scattering fly-ash particles by strongly absorbing have been reviewed. 272 The state-of-the-art model for
soot particles. The addition of soot to coal +fly-ash predicting radiation heat transfer in a complicated
+ gas mixtures (c + f + g) simply decreases the radiat- boiler combustion furnace is based on advanced
ive flux on the cylindrical wall since a greater Monte Carlo type techniques. The model is des-
fraction of the radiant energy is being absorbed by cribed in more detail elsewhere together with exam-
the medium itself. It should be mentioned, however, ples of its practical application. 272 It is shown how
that the effects predicted 6'2~4 in this way may be pilot plant-scale results can be scaled up with the
exaggerated since in these calculations the energy help of the model to predict full-scale performance of
equation is not solved. When radiative transfer is particular boiler furnaces. The uncertainties in pre-
taken into account in the energy equation, the dicting temperatures and heat fluxes are also dis-
temperature would change in a manner that would cussed. It is pointed out that for pulverized coal-fired
partially compensate for the effects of changes in boilers major uncertainties are caused by the un-
radiative properties. known slagging and fouling patterns in the furnace,
The results of sensitivity studies 214 have shown and an ash deposition model could help to reduce
that accurate knowledge of number density, temper- these uncertainties.
ature and particle concentration distributions are Recently, Fiveland and Wesse1298 have developed
more critical than the detailed information about the a very detailed and extensive computer model to
index of refraction of particles and gas concentration predict the performance of three-dimensional pulver-
distributions. The type of coal used affects radiative ized coal-fired furnaces. They have accounted for
transfer relatively little; however, the neglect of fly- almost all of the important physical phenomena that
ash outside of the flame zone has been shown to have can be expected in such systems, including turbu-
a potential for large errors. Apparently, the accuracy lence, chemical reactions, devolatilization, char oxi-
of radiative transfer predictions is not only limited dation as well as radiation heat transfer. Although
by the solution techniques of the radiative transfer they have considered different size particles (e.g.
equation or the prediction of radiative properties, polydispersions) and evaluated the radiative proper-
but mostly by the accuracy of particle concentration ties of particles from Mie theory, scattering in the
and combustion product temperature distributions medium has been considered isotropic. The combus-
which are more time-consuming to evaluate in the tion gas properties have been obtained using the
needed detail. Edwards wide-band model, a5 and the average
The importance of the spatial distribution of properties of the gas-particle mixture have been
radiative properties of pulverized-coal and fly-ash in calculated using the averaging technique proposed
predicting radiation heat transfer accurately was also by Wessel. '26 The radiative transfer equation has
shown by Lowe e t a / . 3 In their analysis they been solved using the discrete transfer method of
employed Hottel's zonal method to solve for radi- Lockwood and Shah; 23"24 however, the method
ative transfer in a utility type pulverized, coal-fired has been revised first to avoid arbitrary radiative
furnace. They showed that furnace heat transfer was source/sink terms encountered in certain volume
insensitive to the type of coal and coal fineness and elements due to numerical diffusion. Wall emissivity
concluded that combustion data were adequate for and thermal conductance of ash deposits can provide
calculation of radiative heat transfer. Lowe et al. 3 a major resistance to heat transfer from the flame-
recommended research on ignition, combustion combustion products to the walls of the furnace, and
stability and radiative properties of fly-ash. these factors were accounted for in the analysis. Flow
Radiation heat transfer 149

FIG. 39. Heat flux isopleths on furnace walls (in W/m2). 29s

patterns, gas temperature, concentration and heat modifications in the radiation model would definitely
flux distributions have been predicted. In Fig. 39 the improve its reliability.
heat flux distribution on the walls of the furnace is
depicted. Note that this figure shows the furnace as
6.3. Gas Turbine Combustors
unfolded. These types of results can be helpful in
identifying potential slagging/fouling problems on It is well established that in gas turbine combus-
membrane walls or convection-pass elements. tors a large fraction of the heat transferred from the
Models of this type are essential to understand the gases to the liner walls is by radiation. The radiation
complex, large-scale, pulverized coal-fired furnaces is due to two contributions: (1) the nonluminous
and are valuable engineering design tools. The radiation emitted by gases such as CO2, H20, CO
radiation heat transfer model needs to be improved and others, and (2) the luminous radiation emitted by
to make it more realistic. Anisotropic scattering by soot particles in the flame. The luminous contri-
particles has been neglected and soot has not been bution from the soot depends on the number and size
taken into account; therefore, the enhancement or of the soot particles. In the primary combustion zone
blockage of radiation by the soot layer is not most of the radiation emanates from the soot
considered. However, as the authors claim, the model particles produced in the fuel-rich regions of the
is still in the initial stages of validation, and further flame. At high pressures encountered in modern
150 R. VISKANTAand M. P. MENGO~:

turbines, the concentrations of soot particles is 1600 I J t

sufficiently large to produce high enough opacities


and consequently soot radiates as a blackbody. It is
under these conditions that radiant heating of the
liner walls is most severe and poses serious problems
12oo
to liner'durability.35
An excellent up-to-date review of radiation heat
transfer from the flame in gas turbine combustors has R~
been prepared. 6 Methods for estimating nonluminous
radiation together with various analytical (global)
models for flame radiation in enclosures are discussed, ~ , ,\//
but attention is focused mainly on the factors that
govern total radiation heat transfer to the liner wall.
The impact of radiation heat transfer on combustor
design features, combustor operating conditions, fuel
I
composition and fuel spray characteristics are dis-
cussed. A need for better understanding of radiative
transfer to establish realistic models for predicting
local heat flux distribution is emphasized. The 0 ~
0 2 4 6 8
understanding can be useful in developing analytical
tools which may lead to improved liner durability in z/r,
future designs by prescribing optimum arrangements FIG. 40. Effect of fuel type (K-kerosine and R50-fuel blend)
for the quantity and distribution of film-cooling air. and of radial soot concentration distribution on radiation
In turn, this approach can also lead to reductions in heat flux at the cylindrical gas turbine combustor wall II-
the time and cost of liner development. 35 uniform and 2,3-nonuniform radial soot distributions). 3""
The simple, global methods based on the mean-
bcam-length concept for predicting flame emissivity
reviewed 6 are not capable of predicting local radi- soot profiles. In practice, such a nearly uniform soot
ation heat flux distribution along the liner wall. concentration profile, though unlikely, might come
Furthermore, simple methods cannot account prop- about if film cooling air of the combustor penetrated
erly for radial and axial nonuniformities of temper- into the combustion zone sufficiently to quench the
ature, species concentration and radiative properties soot oxidation process.
of the soot-gas mixtures. This is a serious short- The results suggest that accurate calculation of the
coming because the combustor designer allocates film radiation heat flux at the combustor wall would
cooling air based on the total heat flux at the liner require both the radial temperature and soot concen-
wall. In the absence of reliable heat flux predictions, tration distributions in the products. Indeed, the
the designer must overprotect the liner. Too much radial temperature distribution had greater impact
cool air near the walls, however, can reduce combus- on the total radiative heat flux than the type of fuel
tion efficiency, increase pollutant emissions, and for the conditions examined in the study. 36 However,
distort the temperature pattern at the combustor scattering of radiation by fuel droplets in a gas
outlet, which stresses the turbine blades. turbine combustor was found to be negligible in
The local radiative flux distributions at the liner comparison to absorption by soot. The average
wall of a typical gas turbine combustor have been radiative heat flux calculated by the P3-approxi-
predicted using the Ps-approximation for radiation marion compared reasonably well with results based
transfer. 36 The mean temperature and soot concen- on the mean-beam-length calculations used in the gas
tration distributions along the combustor were based turbine combustor industry. 6"as However, the P3-
on experimental data. 37 The effects of axial and model results were able to pinpoint locations of
radial temperature and soot concentration distri- maximum radiative flux at the liner wall.
butions, type of fuel, and scattering by fuel droplets The problem of three-dimensional two-phase com-
were investigated. It was found that the axial and bustion has been approached with the aim of
radial temperature and soot concentration distri- producing an algorithm based on fundamental
butions impacted the local radiative flux along the principles which correlate all of the details of
liner wall in several ways. In Fig. 40, the radiative combustion occurring within a gas turbine combus-
fluxes to the cylindrical wall calculated for radially tion can. "~'3s'39 A mathematical model of the
uniform (solid lines) and radially nonuniform (dashed three-dimensional, two-phase reacting flows in gas
lines) soot concentration distributions are compared. turbine combustors has been developed which takes
The medium with a uniform radial soot concen- into account the mass, momentum, and energy
tration yielded larger radiative flux at the liner walls, couplings between the phases, The model incorpor-
at peak, than the nonuniform profile. The temper- ates an accurate representation of the droplet
ature distribution was assumed uniform for both distributions encountered in gas turbine combustors,
Radiation heat transfer 151

and solves the relevant equations for the trajectory 12


Pu
and evaporation of droplets numerically in a
Lo - e p,, / _ ~ _
Lagrangian frame of reference, using a finite-
difference solution of the governing equations of the
gas. Radiative transfer is modeled using the six-flux
approximation, but information on the radiative
properties of the combustion products used in the ~O.G // ~', onstan!
calculations is not provided. The emphasis in the
results reported is on flow and combustion para-
meters as no results on radiative transfer are given. o.z. ~f ....#."

6.4. Internal Combustion Engines "~O" O" I0" 20"


CA
Radiation heat transfer in diesel engines is domin-
ated by the continuum radiation emission by soot FIG. 41. Comparison of total radiation heat losses to a
particles, which are present during the combustion diesel engine cyJinder wall as a function of crank angle
(CA)..~13
process. Radiation also occurs from the carbon
dioxide and water vapor molecules, but because that
energy is concentrated in spectral bands rather than heat flux to the walls adjacent to thin gas zones by as
over the entire spectrum its effect is subordinate with much as 1 0 0 ~ . 313
respect to the energy emitted by the soot. Radiation The most important advantage of differential
is also emitted in bands by many of the intermediate models (like the spherical harmonics approximation)
species formed during combustion, but their effect is is their flexibility to allow for variation of radiative
assumed to be even less important. properties within the medium. In Fig. 41 the total
In spark-ignition engines, where the combustion is radiative flux to diesel engine walls is compared at
usually soot free, the radiation heat transfer is always different crank angles for constant and spatially
small compared to the convection heat transfer. The varying extinction coefficient distributions 3n 3 which
same seems to be the case in diesel engines during were obtained from published experimental data. 3~4
those times in the cycle when soot is not prevalent. It is clear from this figure that using a mean
During combustion the radiation heat transfer is of extinction coefficient to simplify the radiative trans-
the same order of magnitude as the convection heat fer calculations can not always be justified, as the
transfer; whether 25, 50 or 150y,, of the convection fluxes may be underpredicted by about a factor of
heat transfer is a point argued about even in the three. The radiation from soot has been found to be
current literature. The arguments stem from the facts much stronger than that from the gases. 3~J In
that (1) unequivocal heat transfer measurements are addition, the spectral results also reveal distinct
not possible, and (2) the relative importance of spectral selectivity due to the strong gas radiation
convection compared to radiation is highly depen- bands of CO2 and H 2 0 at elevated pressures.
dent upon the engine design and operating character- As in gas turbines, scattering of radiation by fuel
istics. In ceramic-lined engines the convection heat droplets in diesel engines was also found to be
transfer is expected to be reduced more than the negligible compared to absorption by soot. 3z 3 Use of
radiation heat transfer, and thus radiation will be an average homogeneous (position independent)
relatively more important than convection. absorption coefficient in the engine to simplify
Parametric studies of radiation heat transfer in radiation calculations was found to be unjustifi-
diesel engines have been recently reported, an-an3 able. 3~a It was also shown that the distribution of
The method developed by Chang et aL 3j'3~n calcu- radiative flux at the head and piston was incorrectly
lates spectral and total intensity at the chamber walls. predicted and that the total heat loss could be
It is based on the integral form of the RTE along the underpredicted by as much as 60 ~o.
line-of-sight and uses in-cylinder species and temper-
ature distributions as well as a coordinate transfor-
6.5. Fires as Combustion Systems
mation to aid in the integrations. The method is
incompatible with the finite-difference combustion Flame radiation plays an important role in the
models, but can yield accurate results for radiative flame structure, spread and heat transfer from
transfer along the line-of-sight. Spherical harmonics unwanted fires. A recent review 7 has focused on basic
(P~- and P3-) approximations have also been applied aspects of fire and has presented an elementary but
to predict radiation heat transfer for the conditions unified treatment of the phenomenon by considering
encountered in a diesel engine. It has been shown both urban and wildland fires. Several other re-
that the P~-approximation is computationally very views 31"43'239'240'315'316 have treated aspects of
cost effective in comparison to the P3-approximation, flame radiation and have contributed greatly to the
although it overpredicts the total radiation heat loss phenomenoiogy. The interested reader is referred to
to the engine walls by 20 ~ and the local radiation these reviews for books and original research papers
152 R. VISKANTAand M. P. MENGOg:

in the field, and the special issues of Combustion modeling of fire phenomena. Several studies are
Science and Technology (Vol. 39, Nos. 1--6 and Vol. mentioned here.
40, Nos. 1-4, 1984) on Fire Science for Fire Safety, Cooper studied fires in enclosures and described
honoring Professor Howard W. Emmons in which the ceiling jet resulting from the fire,32 the effect of
numerous papers concerned with fires are included. buoyant source in stratified layers, 32~ and the effect
It has now been accepted that radiation is the of side walls in growing fires. 322 However, only in the
dominant mode of heat transfer in fires of large scale, last paper did he consider the effect of radiation
whereas convection (or conductionI is the dominant using simple expressions for radiative transfer to
mode of heat transfer of very small scale fires. estimate the wall temperature. Bagnaro et al. 323
Detailed heat transfer measurements have demon- developed a model to predict experimental room fires
strated that radiation heat transfer from fuel surfaces under steady and transient conditions. They used a
typically exceeds free convection heat transfer for moment method ~77 to solve the radiative transfer
characteristic fuel lengths greater than 0.2 m. 239 equation in three-dimensional enclosures. To repre-
Nonluminous and luminous radiation from turbulent sent the combustion gas contribution they employed
diffusion flames has been recently discussed and a sum-of-gray-gases model. Their results showed
the importance of turbulence/radiation interactions good agreement with experimental data. Also,
has been recently pointed out by Faeth et Markatos and Pericleous 324 studied the effect of
al. 31'254'255 During the last decade there have been radiation on fires in three-dimensional enclosures.
numerous contributions to the literature concerned They employed the six-flux model of Spalding (see
with radiation heat transfer in fires, and it is not Subsection 4.4.1) for the solution of RTE. However,
possible to do justice to them in this very short in neither of these studies is the dependence of the
account. radiative properties on the position (i.e. concen-
Buoyant enclosure flows have applications to tration and temperature) in the medium considered
furnaces and in such phenomena as fire spread in in detail.
rooms and buildings. Numerical and experimental Tien and Lee 43 have provided a comprehensive
studies of two-dimensional and three-dimensional summary of the radiative properties of nonhomo-
turbulent buoyant, simple and complex enclosures geneous and particulate containing media typical of
have been summarized by Yang and Lloyd. 317 The the flame environment. These data can then be used
results obtained have demonstrated that first- in radiation-energy transfer models, which, in turn,
principle numerical finite-difference calculations, determine the characteristics of ignition and fire
together with a simple, yet rational algebraic turbu- spread for the condensed fuel. T M 6.325 331 During
-

lence model, can provide reasonable predictions to a the combustion of condensed fuels, pyrolysis at the
variety of buoyancy-driven vented enclosure-flow fuel surface produces numerous and varied hydro-
phenomena when compared to corresponding experi- carbon gases and soot. The fuel vapors diffuse to the
mental data. The geometries considered unvented flame zone where they react exothermically with
and vented enclosures, aircraft cabin compartments oxygen diffusing from the other side of the flame
and others, but the effects of radiation were neglected. zone. Energy released from the flame zone heats the
At higher temperatures thermal radiation gener- fuel surface, thus maintaining the existing pyrolysis,
ally plays a significant role in affecting the heat creating new areas of pyrolysis, and spreading the
transfer in enclosures such as rooms and buildings, fire. The pyrolyzed gases absorb energy in the
and interactions between thermal radiation and infrared and attenuate the feedback radiation to the
natural or mixed convection must be accounted for fuel surface. This feedback mechanism becomes
in the description of the pertinent momentum and important when the gases are strongly absorbing and
energy transfer processes. Recent discussions on are sooty or when the pathlength becomes large, as in
numerical modeling of natural convection-radiation large-scale fires. For solid and liquid fires, the
interactions in multidimensional enclosures are combustion rate is controlled by the heat transfer
available. 3~a'319 The interactions depend on the from the combustion zone to the fuel surface. In
radiative properties of the absorbing, emitting and large-scale fires (L>0.7 m) fire energy is dominated
scattering media filling the enclosure, a method of by radiation,Qnd the combustion rate is controlled
calculating multidimensional radiative transfer and by radiant feedback from the flame to the fuel
the numerical solution of the governing equations surface. Blockage effects by the pyrolized gases and
for buoyant flows. Current knowledge in these sub- particulates near the fuel surface (discussed in
areas has been discussed. On the basis of these Section 5.2) can attenuate significantly the incoming
reviews,3~ s.319 it is apparent that natural convection- radiation flux. Current analytical models for pre-
radiation interactions in buoyant enclosure flows are dicting the radiation heat flux to the fuel surface
still in the developing stage. An efficient overall consistently overpredict the pyrolysis rate because
computational scheme is still lacking, and metho- the blockage effect is not accounted for. The
dologies which have been developed for natural- assumption of an isothermal and homogeneous flame
convection interaction studies do not appear to have for large scale fires may also lead to significant errors.
been applied to gain improved understanding or The lack of radiative property data for radiation
Radiation heat transfer 153

heat transfer calculations is a major limitation in "'long range" or "action at a distance" transport
improving current fire models. Radiative properties process. In many physical situations radiation can be
of common combustion gases and optical constants modeled without detailed input of complex chemistry,
for soot and simple calculation schemes for deter- chemically reacting turbulent flow and knowledge of
mining the emission coefficients of luminous flames the flame and the reaction region.
have been reviewed. 43 The properties for some of the This review has concentrated on radiation heat
hydrocarbon gas species which are evolved by the transfer in combustion systems. It is clear from the
pyrolysis of condensed fuels, such as plastics, have review that radiation from flames and combustion
been published recently. 332- 33,, Radiative properties products requires detailed information on the radi-
of such gas species as ethylene (C2H4), ethane (C2H,), ative properties of the combustion gases and partic-
propane (C3Hs), methylmethacrylate (C3HsO2), and ulates. Despite the many efforts which have been
others which are major species in pyrolized gases are devoted to the problem, the methods developed for
needed. The wide-band 35 and super-band 4'~ model radiation heat transfer in multidimensional geo-
parameters need to be generated from experimental metries are far from satisfactory, particularly when
data for the radiatively important gases. Total temperatures and gas partial pressures and partic-
emissivity charts can be developed for each gas once ulate concentrations are varying along the path
the band parameters have been determined. These length. The calculation of radiation in combustion
charts graphically express the dependence of total systems is quite involved, and most of the techniques,
emissivity on the temperature, pressure, and optical except those which are called flux or differential
pathlength of the emitting gas and greatly simplify approximations, are incompatible with the numer-
the calculation of flame radiation problems. How- ical algorithms for solving the fluid dynamics-
ever, band information becomes necessary when transport equations.
different gases are combined which have overlapping During the course of the review, a number of
bands in order to determine the correction. Pre- problem areas have been identified and are discussed
dictions of radiative transfer in large-scale fires based in the article. Some specific recommendations for
on data from small-scale flames in laboratory work in modeling radiative transfer in combustion
experiments, however, have been very limited in systems are the following:
accuracy and require much more research attention.
The turbulence/radiation interactions and coupled (1) Radiative property data of less common gases
effects of radiation and flame structure for small such as ethylene (C2H4), ethane (C2H~,), as well
laboratory flames were discussed in Section 5.7. They as propane (C3Hs) and other more important
were found to be more important for luminous than radicals are needed. Radiative properties of
for nonluminous flames. Since smoke (soot) is particulates encountered in pulverized coal com-
generated in open, compartment and building fires bustion such as fly-ash, char and others need to
which are much larger in scale than small laboratory be predicted and verified experimentally. There
flames, the turbulence/radiation interactions are is a very large uncertainty in the radiative
expected to be even more significant because of the properties of these types of particulates that
large and highly variable local opacities that may be have been reported in the literature. Most of the
encountered in these types of systems. The buoyant properties of particles have been obtained at
smoke plume generated by a large fire also involves conditions much different than those encoun-
radiation exchange within itself and with its environ- tered in flames; therefore, it is still not clear
ment. The heat and particulates released by a fire whether these data can be used with confidence
create complex flow patterns which are determined for combustion studies.
by a variety of factors. The interactions of radiation, (2) There has been progress in modeling the thermal
turbulence and flow structure as well as the feedback radiation properties of gases and particulates.
between them in large fires are topics which have However, more research effort is needed, es-
received practically no research attention and are not pecially on physically and analytically well-
understood. founded representations that are simple and
convenient for use in computer codes of com-
7. C O N C L U D I N G REMARKS
bustion systems. Considering that a character-
istic length is always required for use in the
By highlighting recent developments in modeling models and that such a length can not be
radiative transfer, the present review aims to increase rigorously defined for most practical multi-
recognition that very often radiation plays an dimensional systems, it is clear that the concept
important, if not the dominant, role in heat transfer needs additional research attention.
not only in large and intermediate but also in small (3) The nongray effects have been recognized as
combustion systems. Neglect of radiation cannot being very important and it is known that the
be justified in modeling combustion phenomena. gray approximation overpredicts the emission
Modeling of radiative transfer in combustion sys- of radiation from flames with low soot content.
tems can be rather "forgiving" because radiation is a The calculations of radiative transfer for non-
JPEC8 13 : 2 - g
154 R. VISKANTAand M. P. MENG0~

homogeneous, nonisothermai flames on a non- (8) Research effort should be devoted to experiment-
gray basis would enable accurate predictions of ally validating the radiative transfer model(s) in
flame emission for a wide range of pathlengths~ order to demonstrate the potential usefulness of
The results could then be used to establish the methods to the analysis and design of
scaling relations and to assess the range of practical systems.
validity of the gray analysis.
(4) In combustion systems involving the burning of Acknowledgements Much of the author's recent work
solid fuels such as pulverized coal, the particles reported in this review was supported by CONOCO. Inc.
and gases surrounding them are at different through a grant to the Coal Research Center of Purdue
University. It is a pleasure to acknowledge CONOCO's
temperatures. Analytical models based on interest in fundamental radiation heat transfer research
experiment need to be developed to predict rehlted to combustion systems. The authors wish to express
radiative transfer and temperatures in such their appreciation to Miss Nancy Rowe for her dedicated
systems. The slip between particles and gases help in transforming their notes into a polished manuscript.
The authors are also indebted to the anonymous reviewers
must be considered. This is not only important
for pointing out typographical errors and for suggesting
for predicting accurately the flow and temper- improvements in the presentation.
ature fields, but also necessary for the under-
standing of soot formation and soot volume
fraction distribution in the medium.
(5) In most practical, large-scale combustion sys- REFERENCES
tems the chemically reacting flow is turbulent.
The question needing an answer is to what 1. ORAN, E. S. and Bo)us, J. P., Prog. Energy Combust.
Sci. 7, 1 (1981).
extent the interaction of turbulence and radi- 2. SMOOT, L. D., Prog. Energy Combust. Sci., I0, 229
ation will modify the flow properties, radiative (1984).
transfer and temperature in the combustion 3. SMOOT, D. L. and SMITH, P. J., Coal Combustion and
system. In turn, this will affect the chemical Gasification, Plenum Press.New York (1985).
reactions, radiating species concentrations and 4. BLOKH, A. G., Heat TransJer in Steam Boiler Furnaces,
Energoatomizdat, Leningrad (1984) (in Russian) (to be
their distributions as well as the flame structure. published by Hemisphere Publishing Corp., Washing-
This may be particularly important in large- ton, D.C.).
scale, highly turbulent flames and fires con- 5. KHALIL, E. E., Modeling Furnaces and Combustors,
taining soot. Abacus Press, Tunbridge Wells, Kent (1982).
6. LEFEBVRE,A. H., Int. J. Heat Mass Tran.~fer 27, 1493
(6) Rigorous and relatively simple models for (1984).
handling radiative transfer in one-dimensional 7. WILLIAMS, F. A., Prog. Energy Combust. Sci, 8, 317
and some two-dimensional geometries are avail- (1982).
able; however, there is still a need for effective, 8. Kuo, K. K., Principles of Combustion, Wiley, New
accurate, and simple-to-use multidimensional York (1986).
9. BIRD, R. B., STEWART. W. E. and LIGHTFOOT, E. N..
models. Radiative transfer should be accounted Transport Phenomen,:, Wiley, New York (1960).
for in the thermal energy equation when model- 10. SAMPSON,D. H., Radiative Contributions to Energy and
ing combustion phenomena. The accuracy of the Momentum Transport in a Gas, Interscience Pub-
radiation model should be compatible with that lishers, New York (1965).
11. VINCENn, W. G. and KRUGER, JR., C. H., Introduction
of the combustion model. The calculations
to Physical Gas Dynamics, Wiley, New York (1965).
should be interactive in nature, that is, radiative 12. PREISENIX)RFER, R. W., Radiative Transfer Theory on
properties should be predicted from the know- Discrete Spaces, Pergamon Press, New York (1965).
ledge of the gas and particle concentrations, and 13. HARMS, E. G. and SIMON, A., Physics Fluids 3, 255
these properties should then be used in calcu- (1960).
14. OSBORN, R. K. and KLEVANS, E H., Ann. Phys. 15, 105
lating local radiation heat transfer, temperature (1961).
distributions and local radiating species concen- 15. CHANDRASEKHAR,S., Radiative TransJer, Oxford Uni-
trations. Because of the nonlinearities of the versity Press, London (1950)(also Dover Publications,
processes, such calculations will, most likely, New York (1960)).
16. KOURGANOFF,V., Basic Methods in TransJer Problems,
have to be carried out iteratively. Oxford University Press, London (1952).
(7) Effort should be devoted to develop approxi- 17. SOBOLEV,V. V., A Treatise on Radiative Tran,~fer, D.
mate, but physically sound, relations Claws") for Van Nostrand, Princeton, N.J. (1963).
scaling radiative transfer in combustion systems. 18. BORN, M. and WOLF E., Principles of Optics, 2nd Edn,
Such relations are needed for scaling small Pergamon Press, Oxford (1964).
19. OZi~lg, M. N., Radiative Transfer and Interactions with
laboratory flames (combustion systems) to large Conduction and Convection, Wiley, New York (1973).
scale ones typical of real or practical combus- 20. SIEGEL,R. and HOWELL, J. R., Thermal Radiation Heat
tion systems. Most likely some of these laws will Transfer, 2nd Edn, Hemisphere, Washington, D.C.
be empirical in nature; therefore, experimental (1981).
21. CARLSON, B. G. and LATHROP, K. D., Computing
data will be needed for small laboratory, proto-
Methods in Reactor Physics, H. Greenspan, C. N.
types as well as full-scale systems to validate the Kelber and D. Okrent (Eds), Gordon and Breach,
relations. Science Publishers Inc., New York (1968).
Radiation heat transfer 155

22. LEDOUX, P. and WALRAVEN,TH., Htmdhuch der Physik, 52. SMrm, T. F., SHEN, Z. F. and FRIEDMAN,J. N., J. Heot
S. Flugge (Ed.), Vol. 11, pp. 353-604, Springer Verl,'tg, Transfer 104. 602 (1982[
Berlin 11958[ 53. FARAG, I. H., Heat Transfer 1982, U. Grigull, E.
23. SIMON, R., J. qu,mtre Spectrosc. radiot. Tron.~/i i, 3 Hahne, K. Stephan and J. Straub (Eds), VoL 2. pp.
11963). 489-492. Hemisphere, Washington, D.C. (1982).
24. TOWNSEND,A. A., J. Fluid Mech. 3, 361 11958). 54. COPALLE,A. and VERVlSCH,P., Combust. Flame 49. 101
25. SHRED, G. M. and AKMAYEV, R. A., Atmos. Ocean. 11983).
Phys. I0, 894 11974). 55. GROSSHANDLER, W. L., Int. J. Heot Mass Tran.~fer 23,
26. PRINN, R. G., J. atmos. Sei. 34, 1386 11977). 1447 11980).
27. PEARCE,B. E. and VARMA,A. K., Radiation-turbulence 56. VISKANTA, R., lnt. J. Heot Moss Tron,~/er 7, 1047
interaction in a tactical missile exhaust plume, AIAA (1964[
Pap. No. AIAA-81-1110 (1981). 57. TRAUGOTT,S. C., J. quontve Spectrosc. radiat. TransJer
28. TAN, E. and FOSTER, P. J., Heat TransJer--1978, 8, 971 11968).
Hemisphere Publishing Corp., Vol. 3, pp. 403-408, 58. PATCH, R. W., J. quantve Speetrosc. radiat. Transfer 7,
Washington, D.C. 11978). 611 11967).
29. FERGUSON, C. R. and MELLOR, A. R., Radiative heat 59. MENGO~, M. P. and VISKANTA, R., Heat Transfer--
transfer from gas turbine flames, ASME Pap. No. 79. 1986, C. L. Tien, V. P. Carey. and J. K. Farrell (Eds),
GT-144 (1979). Vol. 2, pp. 815-820, Hemisphere Publishing Corp.,
30. SONG. T. H. and VISKANTA, R., J. Thermophys. Heat Washington, D.C. 11986).
Transfer 1, 56 0987). 60. FELSKE,J. D. and TIEN, C. L., Combust. Sei. Teehnol. 7,
31. FAE-IrH,G. M., JE'NG, S.-M. and GORE, J., Heot Tron.~/br 25 11973).
in Fire and Combustion Systems, C. K. Law, Y. Jaluria. 61. MENGO~:, M. P. and VISKANTA, R., J. Heat TransJer
W. W. Yuen and K. Miyasaka (Eds), pp. 137-151, 108, 271 11986),
ASME, New York (1985). 62. CROSBII~A. L. and VISKANTA,R., J. qutmtt'e Spectrosc.
32. GOODY, R. M., Atmospheric RadiatiotL Oxford Univer- radiat. TronsJ~,r 9, 533 (1969).
sity Press, London 11964). 63. BEER,J. M.. J. inst. Fuel 37, 286 11964).
33. TIL~,~,C. L., Adranees in Heat Tran.~/er, T. F. Irvine, Jr. 64. SAROFIM, A. F. and HOTTEL, H. C., Heat TronsJer--
and J. P. Hartnett (Eds), Vol. 5, pp. 254-324, Academic 1978, Vol. 6, pp. 199-217, Hemisphere Publishing
Press, New York (1968). Corp., Washington, D.C. 11978).
34. LrdDWIG. C. B,, MALKMUS, W., REARDON, J. G. and 65. VAN DE HULST, H. C., Light Scattering by Small
THOMSON, J. A. L., Handbook o/inJi'ared Radiation Particles, Wiley, New York 11957) (also Dover, New
From Combustion Gases, R. Goulard and J. A. L. York 11981)).
Thomson (Eds), NASA SP-3080, Washington, D.C. 66. KERKER, M.. The Scattering of Light, Academic Press,
11973). New York 11969).
35. EDWARDS, D. K., Adranees in Heat TransJer, T. F. 6% FLAGAN, R, C., Seventeenth Symposium (International)
Irvine, Jr and J. P. Hartnett (Eds), Vol. 12, pp. 115-193, on Combustion, pp. 97-104, The Combustion Institute,
Academic Press, New York (1976). Pittsburgh, PA 11979).
36. EDWARDS,D. K. and MENARD, W. A., Appl. Opt. 3, 621 68. CHYLEK,P., GRAMS,G. W. and PINNIeK, R. G., Science
(1964). 193, 48011976).
37. TIES, C. L. and LOWDER, J. E., Int. J. Mass Tran.~fer 9, 69, FIELD, M. A., GILL, D. W., MORGAN, B. B. and
698 11966). HAWKSLEY, P. G. W., Combustion of Pulveri=ed Coal,
38. Hsmn, T, C. and GRnF, R., Int. J, Heat Mass Trons.]t,r The British Coal Utilization Research Association,
15, 1477 11972). Leatherland (I 967).
39. EDWARDS, D. K. and BALAKRISHNAN,A., Int. J. Heat 70. DALZELL, W. H. and SAROFIM,A. F., J. Heat TransJer
Mass Transfer 16, 25 11973). 91,100 (1969).
40. FELSKE, J. D. and TIEN, C. L., int. J. Heat Mass 71. LEE, S. C. and TIEN. C. L., Eighteenth Symposium
Trtmsfer 17, 155 11974). (International) on Combustion. pp. 1159-1166, The
41. TIWARI, S. N., Int. J. Heat Mass Transfer 20, 741 Combustion Institute, Pittsburgh, PA (1980).
(1977). 72. BARD, S. and PAGNI, P. J., J. Heat Transfer 103, 357
42. TIWARI, S. N., Adwmees in Geophysics, B. Saltzman 11981).
(Ed.), Vol. 20, pp. 1-85, Academic Press, New York 73. FELSKE, J. D., CHARALAMPOPOULOS,T. T. and HURA,
11978). H. S., Combust. Sci. Teehnol. 37, 263 (1984).
43. T~EN,C. L. and LEE,S. C., Prog. Energy Combust. SoL 8, 74. FELSKE, J. D., HSU, P.-F. and KtJ, J. C., J. quontre
41 (1982). Spectrose. radiat. Transfer 35. 447 11986).
44. Yu, Q. Z., BROSMER,M. A. and TIEN, C. L., Proceedings 75. FOSTER, P. J. and HOWARm, C. R., Carbon 6, 719
of the 1985 International Symposium on Heat Transfer, 11968).
Beijing, China 11985) (to appear). 76. BREWSTER, M. Q. and KtmITOMO, T., Proceedings of
45. MI~GO~:, M. P., Modeling of radiative heat transfer in ASM E - J S M E Thermal Engineering Joint Confi, rence.
multidimensional enclosures using spherical har- Vol. 4, pp. 21-26, ASME/JSME, New York 11983).
monics approximation, Ph.D. Thesis, Purdue Univer- 77. WYATT, P. J., Appl. Opt. 19, 975 11980).
sity, West Lafayette, IN (1985). 78. WALL,T. F., LOWE, A., WmnERLEY, L. J. and STEWART,
46. HOTTEL,H. (2. and SAROFIM,A. F., Radiatire Tran.~/er, !. McC., Prog. Energy Combust. Sci. 5, 1 (1979).
McGraw Hill, New York 11967). 79. Low~,A., Srt~V^RT,I. McC. and WALL,T. F., Seventeenth
47. EDWARDS, D. K. and MATAVOSIAN, R., J. Heat Symposium {International) on Combustion, pp. 105-114,
TransJer 106, 684 (1984). The Combustion Institute, Pittsburgh, PA 11979).
48. LECKNER,B., Combust. Flame 19, 33 (1972). 80. GROSSHANDLER,W. L. and MONT~RO, S. L. P., J. Heat
49. MODAK, A. T., Fire Res. I, 339 11979). TransJer 104, 587 (1982).
50. STEWARD, F R. and KOCAEFE, Y. S., Heat TronsJer-- 81. GOODWIN,D. G. and MITCHNER, M., Measurements of
1986, C. L. Tien, V. P. Carey, and J. K. Ferrell (Eds), the near infrared optical properties of coal ash, ASME
Vol. 2, pp. 735-740, Hemisphere Publishing Corp., Paper No. 84-HT-41 (1984[
Washington, D.C. (1986). 82. GUPTA, R. P. and WALL,T. F., Combust Flame61, 145
51. TAYLOR, P. B. and FOSTER, P. J., Int. J. Heat Mass (1985[
TransJer 17, 1591 11974). 83. GOODWlN, D. G., Infrared Optical Constants of Coal
156 R. VISKANTAand M. P. MENG0~

Slags, Topical Rep. No. T-255, High Temperature Gas 119. HUFFMAN,D. R. and BOHREN,C. F., Light Scatterinq by
Dynamics Laboratory, Mechanical Engineering Irregularly Shaped Particles, D. W. Schuerman led.),
Department, Stanford University, Stanford, CA (1986). pp. 103-11 I, Plenum Press. New York (1980).
84. CHENG, R. J., Light Scattering by Irregularly Shaped 120. CHu, C. M. and CHURCHILL, S. W., J. Ol't. Soc. Am. 45,
Particles, D. W. Schuerman ted.), pp. 69-78, Plenum 958 (1955).
Press, New York (1980). 121. MCKELLAR, B. H. J. and Box, M. A,, J. atmos. Sci. 38,
85. WIRBERLEY, U J. and WALL, T. F., Combust. Sci. 1063 (1981).
Technol. 48, 177 (1986). 122. JOSEPH, J. H., WISCOMBE, W. J. and WEINMAN,J. A., d.
86. MCLEAN, W. J., HARDESTY, D. R. and POHL, J. H., atmos. Svi. 33, 2452 (1976).
Eighteenth Symposium (International) on Combustion, 123. CRUStaL, A. L. and DAVIDSON, G. W., J. quantve
pp. 1239-1248, The Combustion Institute, Pittsburgh, Spectrosc. radiat. Transfi'r 33, 391 (1985).
PA (1981). 124. CLARK, G. C., CHU, C. M. and CHURCHILL, S. W., J.
87. SEEKER, W. R., SAMUELSEN, G. S., HEAP, M. P. and opt. Sot'. Am. 47, 81 (1957).
TROLINGER, J. D., Ei~jhteenth Symposium (International) 125. BREWSTER, M. Q. and TIEN, C. L., Int J. Heat Mass
on Comh,stion, pp. 1213-1226, The Combustion TransJer 25, 1905 (1982).
lnstitue, Pittsburgh, PA (I 981 ). 126. WESSEL, R. A., Heat TransJer in Fire and Combl~stion
88. WISCOMBE, W. J. and MUGNAI, A., Single Scatterimj Systems, C. K. Law, Y. Jaluria, W. W. Yuen and K.
Ji'om Nonspherical Chebyshev Particles: A Compendium Miyasaka (Eds), pp 239-249, ASME, New York (1985).
of Calculations, NASA Reference Publication 1157, 127. SCrtOLAND, E. and SCHENKEL, P., Heat TransJer--
Washington, D.C. (1986). 1986, C. L. Tien, V. P. Carey, and J. K. Ferrell lEd.),
89. MIE, G., Ann. Phys. 25, 377 (1908). Vol. 2, pp 763-768, Hemisphere Publishing Corp.,
90. BOHREN, C. F. and HUEFMAN, D. R., Absorption and Washington, D.C. (1986).
Scatterimj of Light by Small Particles, Wiley, New 128. CAR'nGNV,J. D., Heat Transfer--1986, C. U Tien, V.
York (1983). P, Carey and J. K. Ferrell (Eds), Vol. 2, pp. 769-772,
91. WAIT, J., Can. J. Phys. 33, 189 (1955). Hemisphere Publishing Corp., Washington, D.C.
92. LIOU, K., Appl. Opt. 11,667 (|972). (1986).
93. COHEN, A. and ALPER, P., Appl. Opt. 18, 2738 11979). 129. SKOCYPEC,R. D. and BUCKIUS, R. 0., Int. J. Heat Mass
94. COHEN, A., HARACZ, R. D. and COHEN, L. D., J. appl. TransJ~,r 2% 1 11984).
Phys. 58, 1135 (1985). 130. SKOCYPE,C, R. D., WALTERS, D. V. and BUCKIUS, R. O..
95. YEn, C., J. opt. Sot'. Am. 55, 39011965). Comhust. Sci. Tevhnol. 47, 239 (1986).
96. ASANO,S. and YAMAMOTO,G., AppL Opt. 14, 29 (1975). 131. DAVlSON, B., Neutron Transport Theory, Clarendon,
97. DAVE,J. V., IBM JI Res. 13, 302 (1969). Oxford 11958).
98. WISCOMBE,W., Mie Scattering Calculations: Improve- 132. CASE, K. M. and ZWEIEEL, P. F., Line,n" Transport
ments in Technique and Fast, Vector-Speed Computer Theory, Addison-Wesley, Reading, MA (1967).
Codes, NCAR TN-140 + STR, Boulder, CO (1979). 133. LEWIS, E. E. and MILLER, JR., W. F., Comp,tational
99. WISCOMBE,W., Appl. Opt. 19, 1505 (1980). Methods t~] Neutron Transport, Wiley, New York
100. ME(, K., MORGAN, M. and CHANG,S., Electromagnetic (1984).
Scatterimj, P. Uslenghi ted.), pp. 359-392, Academic 134. VISKANTA,R., Advances in Hem Tran.~]er, T. F. Irvine
Press, New York (1978). and J. P. Hartnett (Eds), Vol. 3, pp. 175-251, Academic
101. YEn, C. and MEI, K., Light Scattering./?ore Irregularly Press, New York (1966).
Shaped Particles, D. W. Schuerman tEd.), pp. 201-206, 135. CRUSH]E, A. L. and DOUGHERTY, R. L., J. qutmtre
Plenum Press, New York (1980). Spectrosc. radiat. Transfi,r 25, 551 11981).
102. WATERMAN,P., Proc. IEEE 53, 805 (1965). 136. HUNT, G. E., J. Inst. mathl Applic'. 3, 181 (1967).
103. WATERMAN,P., Phys. Rev. D3, 825 (1971). 137. CHENG, P., Thermal Control and Radiation, C. L. Tien
104. BARaER, P. and YEn, C., Appl. Opt. 14, 2864 (1975). tEd), Progress in Astronautics and Aeronautics, pp.
105. KELLER,J., J. opt. Sot'. Am. 52, 116 (1962). 269-308, MIT Press, Cambridge (1973).
106. HARRINGTON, R., Field Computations hy Moment 138. DUA, S. S. and CHE~G, P., Int. J. Heat Mass Tran.~/br
Methods, MacMillan, New York (1968). 18, 245 (1975).
107. YEn, C. Phys. Rev. 135, 1193 (1964). 139. CRUStaL, A. L. and SCHRENKER, R. G., J. quantve
108. OGUCHI, T., Radio Sei. 8, 31 (1973). Spectrosc. radiat. TransJ~,r 20, 507 11982).
109. POLLACK.J. and Cvzza, J., J. atmos. Sci. 37, 868 (1980). 140. CRUSH(E,A. L. and FARRELL,J. B., J. quantve Spectrosc.
110. SCHUERMAN,D. W., Light Scattering from Irregularly radiat. TransJer 31,397 (1984).
Shaped Particles, Plenum Press, New York (1980). 141. CROSR]E, A. L. and SCHRENKER, R. G., J. quantve
111. MUGNAI, A. and W[SCOMaE, J. W., Appl. Opt. 25, 1235 Spectrosc. radiat. Transfer 33, 101 (1985).
(1986). 142. SELgUK,N., J. Heat Transfer 107, 648 (1985).
112. MENGO(;, M. P. and VISKANTA, R., Combust. Sci. 143. MCKEE, T. B. and Cox, S. K., J. atmos. Sci. 31, 1885
Technol. 44, 143 (1985). (1974).
113. BUCKIUS, R. O. and HWANG, D, C., J. Heat TransJer 144. DAVIES,R., J. atmos. Sci. 35, 1712 (1978).
102, 99 (1980). 145. HOWELL,J. R. and PERLMUTTER, M., J. Heat TransJer
114. VISKANTA, R., UNGAN, A. and MENGOq, M. P., 86, 116 (1964).
Prediction of radiative properties of pulverized coal 146. HOWELL,J. R., Advances in Heat TransJer, T. F. lrvine
and fly-ash polydispersions, ASME Pap. No. 81.HT- and J. P. Hartnett (Eds), Vol. 5, pp. 1-54, Academic
24, ASME, New York (1981). Press, New York (1968).
115. BUCKIUS,R. O., Heat TransJer--1986, C. L. Tien, V. P. 147. TAN]GUCHI,H., Bull J.S.M.E. 12, 67 (1969).
Carey and J. K. Ferrell (Eds), Vol. 1, pp. 141-150, 148. STEWARD, F. R. and CANNON, P., Int. J. Heat Mass
Hemisphere Publishing Corp., Washington, D.C. TransJer 14, 245 (1971).
(1986). 149. TANIGUCHI,H., KUDO. K., HAYASAKA,H., YANG, W.-J.
116. STEPHENS,G. L., Appl. Opt. 23. 954 (1984). and TASHIRO, H., Fundamentals of Thermal Radiation
117. MACKOWSKI,D., ALTENK1RCH,R. A. and MENGOq, M. Heat TransJer, T. C. Min and J. L. S. Chen (Eds),
P., Combust. Sci. Technol. (in press). ASME HTD--Vol. 40, pp. 29-36, New York (1984).
118. SIEGEL, R., Radiative behaviour of a gas layer seeded 150. BROWN, F. B. and MARTIN, W. R., ProW. nucl. Energy
with soot, NASA TN D-8278, Washington, D.C. 14, 269 (1984).
11976). 151. HOTTEL, H. C. and CUmiN, E. S., A.I.Ch, E. JI 4, 3
Radiation heat transfer 157

(1958)~ 185. RATZF,L III, A. C. and HOWELL, J. R., J. Heat TransJer


152. VERCAMMEN,H. A. J. and FROMENT,G. F., Int. J. Hetat 105, 333 (1983).
Mass Transfer 23, 328 (1980). 186. LIOU, K. N. and Ou, S. C., J. atmos. Svi. 36, 1985
153. NOnLE, J. J., Int. J. Heat Mass Transjer 18, 261 (1975). (1979).
154. SMITH, T. F., BYUN, K.-H. and FORD, M. J., Heat 187. MENGO~, M. P. and V1SKANTA, R., J. quantve
TransJer--1986, C. L. Tien, V. P. Carey and J. K. Spectrosc. Tad(at. TransJer 33, 533 (1985).
Ferrell (Eds), Vol. 2, pp. 803-808, Hemisphere, 188. SHOKAIR, I. R. and POMRANING, G. C., J. quantve
Washington, D.C. (1986). Spevtrosv. Tad(at. Transfer 25, 325 (1981 ).
155. LARSEN, M. E. and HOWELL, J. R., J. Heat Trans./er 189. YVON,J., J. nucl. Energy I 4, 305 (1957).
107, 936 (1985). 190. SOtIFF, D. and ZIEmNG, S., Nucl. Sci. Engng 7, 172
156. HOWELL, J. R., Radiative transfer in multidimensional (1960).
enclosures with participating media, A S M E Pap. No. 191. OZa,~K, M. N., MENNING, J. and HALG, W., J. quantve
83-HT-32 (1983). Spectrosc. Tad(at. Transfer IS, 1101 (1975).
157. STEWARD, F. R. and TENNANKORE, K. N., J. Inst. 192. UNNO, W. and KONDO, M., Pubis astr. Soc. Japan 28,
Energy LII, 107 (1979). 347 (1976)
158. SIDDALL,R. G., Heat TransJer--1986, C. L. Tien, V. P. 193. WILSON, S. J. and SEN, K. K., J. quantve Spectrosc.
Carey and J. K. Ferrell (Eds), Vol. 2, pp. 751-756, Tad(at. TransJer 35, 467 (1986).
Hemisphere Publishing Corp., Washington, D.C. 194. LATHROP,K. D. and BmNKLEY, F. W., TWOTRAN-II:
(1986). An interface exportable version of the TWOTRAN
159. WHITNEy, C., J. quantve Spectrosc. radiat. Transfer 14, code for two-dimensional transport, Rep. LA-4848-
591 (1974). MS, Los Alamos Scientific Laboratory (1973).
160. KROOK, M., Astrophys. J. 122, 487 (1955). 195. LATHROP,K. D., THREETRAN--A program to solve
161. VISKANTA,R., Fortschr. VeT Tevh. 22A, 51 (1984). the multigroup discrete ordinates transport equation
162. INCROPERA, F. P., CRAIG, T. D. and HOUF, W. G., J. in (x,y,z) geometry, Rep. LA-6333-MS, Los Alamos
quantve Spectrosc. radiat. Transjer 31,139 (1984). Scientific Laboratory (1976).
163. SIDDALL, R. G., J. Inst. Fuel 47, 101 (1974). 196. CARLSON,B. G. and LEE, C. E., Mechanical quadrature
164. MENGOt~, M. P. and VISKANTA, R., J. quantve Spec- and the transport equation, Rep. LA-2573, Los Alamos
trosc. Tad(at. TransJer 29, 381 (1983). Scientific Laboratory, Rep. LA-2573 0961).
165. ABRAMZON,M. N. and LISIN, F. N., High Temp. 22, 95 197. HOUF, W. G. and INCROPERA, F. P., J. quantve
(1984). Spevtrosc. Tad(at. TransJer 23, 101 (1980).
166. SIDDALL,R. G. and SELCUK, N., Trans. I.Ch.E. 57, 163 198. KHALIL, H., SHULTIS, J. K. and LESTER, T. W., Num.
(1979). Heat Transfer 5, 235 (1982).
167. SPALDING,D. B., Personal notes (1970). 199. FIVELAND,W. A., J. Heat TransJer 106, 699 (1984).
168. WHITACRE,G. R. and MCCANN, R. A., Comparison of 200. GERSTL, S. A. W. and ZARDECK, A., Appl. Opt. 24, 81
methods for prediction of radiant heat flux distri- (1985).
bution and temperature, ASME Pap. No. 75-HT-9 201. LATHROP,K. D., Nuvl. Sci. Engng 32, 357 (1968).
(1975). 202. LATHROP,K. D., Nucl. Sci. Engng 45, 255 (1971).
169. GOSMAN, A. D. and LOCKWOOD, F. C., Fourteenth 203. SHAH, N. G., New method of computation of radiant
Symposium (International) on Combustion, pp. 661-671, heat transfer in combustion chambers, Ph.D. Thesis,
The Combustion Institute, Pittsburgh, PA (1973). Imperial College, London, England (1979).
170. RICHTER, W. and QuicK, R., Heat TransJer in Flames, 204. LOCKWOOD, F. C. and SHAH, N. G., Eighteenth
N. H. Afgan and J. M. Beer (Eds), pp. 95-109, Scripta Symposium (International) on Combustion, pp. 1405-
Book Co., Washington, D.C. (1974). 1414, The Combustion Institute, Pittsburgh, PA
171. DOMOTO,G. A. and WANG, W. C., J. Heat TransJer96, (1981).
385 (1974). 205. TANIGUCHI,H., YANG, W.-J., KUDO, K., HAYASAKA,H.,
172. LowEs, T. M., BARTELDS, H., HEMP, M. P., OGUMA, M., KUSAMA,A., NAKAMACHI, I. and OKIGANI,
MICHELFELDER, S. and PAl, B. R., Heat TransJer in N., Heat TransJer--1986, C. L. Tien, V. P. Carey and
Flames, N. H. Afgan and J. M. Beer (Eds), pp. 179-190, J. K. Ferrell (Eds), Vol. 2, pp. 757-762, Hemisphere
Scripta Book Co., Washington, D.C. (1974). Publishing Corp., Washington, D.C. (1986).
173. Cnu, C. M. and CHURCHILL, S. W., J. phys. Chem. 59, 206. EDWARDS, D. K., Heat TransJer in Fire and Combus-
855 (1955). tion Systems, C. K. Law, Y. Jaluria, W. W. Yuen and K.
174. VARMA,S. A., Pulverized Coal Combustion and Gasifi- Miyasaka (Eds), pp. 273-278, ASME, New York
cation, L. D. Smoot and D. T. Pratt (Eds), pp. 311-315, (1985).
Plenum Press, New York (1979). 207. SEL(~UK,N. and SIDDALL, R. G., Int. J. Heat Mass
175. FIVELAND, W. A., A discrete ordinate method for Transfer 19, 313 (1976).
predicting radiative heat transfer in axisymmetric 208. HARSHVARDHAN, WEtNMAN, J. A. and DAVIES, R.,
enclosures, ASM E Pap. No. 82-HT-20 (1982). Transport of infrared radiation in cuboidal clouds,
176. DEMARCO, A. G. and LOCKWOOD, F. C., Riv. Combust. NASA TM-82116 (1981).
29, 184 (1975). 209. PR[]SENDORFER, R. W. and STEPHENS, G. U, J. atmos.
177. LOCKWOOD, F. C. and SHAH, N. G., Heat TransJer-- Sci. 41,709 (1984).
1978, Vol. 2, pp. 33-40, Hemisphere Publishing Corp., 210. STEPHENS,G. L. and PREISENDORFER, R. W., J. atmos.
Washington, D.C. (1978). Svi. 41,725 (1984).
178. JEANS,J. H., Mon. Not. R. astr. Soc. 78, 28 (1917). 211. VAN DE HULST, H. C., Multiple Light Scattering,
179. POMRANING,G. C., Ann. Phys. 27, 193 (1964). Tables, Formulas and Applications, Vol. 1, Academic
180. BAYAZITOg:-LU,Y. and HIGENYI, J., AIAA Jl 14, 424 Press, New York (1980).
(1979). 212. RAZZAQUE,M. U., KLAN, D. E. and HOWELL, J. R., J.
181. FLETCHER,J. K., Nucl. Sci. Enong 84, 33 (1983). Heat Transfer 105, 933 {1983).
182. HIGENYI, J. and BAYAZITO~LU, Y., AIAA Jl 18, 723 213. SELg:UK,N., SIDDALL, R. G. and BEER, J. M., Sixteenth
(1980). Symposium (International) on Combustion, pp. 53-62,
183. HIGENYI, J. and BAYAZITOOLU, Y., J. Heat Trans/er The Combustion Institute, Pittsburgh, PA (1976).
102, 719 (1980). 214. ML~GOg:, M. P. and VISKANTA, R., Combust. Sci.
184. AHLUWALIA, R. K. and Ira, K. H., Int. J. Heat Mass Technol. 51, 51 (1987).
TransJer 24, 1421 (1981). 215. PICCIRELLI, R. A., AaLUWALIA, R. K. and IM, K. H.,
158 R. VISKANTAand M. P. MENGOg:

Heat Transfer in Fire and Combustion Systems, C. K. 247. ABDEL-KHALIK,S., TAMARU.T. and EL-WAKIL, M. M.,
Law, Y. Jaluria, W. W. Yuen and K. Miyasaka (Eds). Heat TransJer in Flames, N. H. Afgan and J. M. Beer
pp. 231-238, ASME, New York (1985). (Eds), pp. 365-373, Scripta Book Co., Washington,
216. LAURENDEAff,N. M., Proud. Energy Combust Sci. 4, 221 D.C. (1974).
(1978). 248. KINOSHITA, C. M. and PAGNI, P. J., Eighteenth
217. LAW, C. K., Proo. Energy Combust. Sci. 8, 171 (1982). Symposium (International) on Combustion, pp. 1415-
218. CHOI, S. and KRUGER, C. H., Combust. Flame 61, 131 1425, The Combustion Institute, Pittsburgh (1980).
(1985). 249. LIu, K. V,, LLOYD, J. R. and YANG, K. T., int. J. Heat
219. VISKANTA,R. and MERRIAM,R. L., J. Heat Transfltr 90, Mass Trans./er 24, 1959 (1981).
248 (1968). 250. JAMES,R. K. and EDWARDS,O. K., J. Heat TransJer 99,
220. GROSSHANDLER,W. I., Combust. Flame 55, 59 (1984). 221 (1977).
221. TIMOTHY, L. D., SAROFIM, A. F. and BEER, J. M., 251. KENT, J. H. and BILGER, R. W., Fourteenth Symposium
Nineteenth Symposium (International) on Combustion, (International) on Combustion, pp. 615-625, The Com-
pp. 1123-1130, The Combustion Institute, Pittsburgh bustion Institute, Pittsburgh, PA (1973).
(1982). 252. BILGER, R. W. and BECK, R. W., Fifteenth Symposium
222. SOTIRCHOS, S. V., Proceedings of the First Annual (International) on Combustion, pp. 541-552, The Com-
Pittsburgh Coal Conference, pp. 176-189, Pittsburgh, bustion Institute, Pittsburgh, PA (1975).
PA, US Department of Energy Rep. CONF 840971 253. STARNER, S. H. and BILGER, R. W., Eighteenth
(DE 85012243) (Sept. 17-24, 1984). Symposium (International) on Combustion, pp. 921-930,
223. SO~RCHOS, S. V. and AMt~DSON, N. R., Ind. Engng The Combustion Institute, Pittsburgh, PA (1981).
Chem. Fundam. 23, 191 (1984). 254. FAETH,G. M., Heat TransJer--1986, C. L. Tien, V. P.
224. SOaaRCHOS, S. V. and AMtmDSON, N. R., A.I.Ch.E. JI Carey and J. K. Farrell (Eds), Vol. 1, pp. 151-160,
30, 537 (1984). Hemisphere Publishing Corp., Washington, D.C.
225. KOOKER, D. G., Combust. Flame 49, 141 (1983). (1986).
226. SOHRAB, S. G. and LAW, C. K., Int. J. Heat Mass 255. FAETH,G. M., GORE, J. P., CHUECH, S. G. and JENG, S.-
TransJer 27, 291 (1984). M., A. Re~. num. Fluid Mech. Heat Transfl,r lin press~
227. LEE, K. Y., ZHONG, Z. Y. and TIEN, C. L., Twentieth 256. GROSSHANDLER,W. L., lnt. J. Heat Mass TransJer 23,
Symposium (International) on Combustion, pp. 1629- 1147 (1980).
1636, The Combustion Institute, Pittsburgh, PA 257. GROSSHANDLER, W. L. and NGUYEN, H. D., J. Heat
(1984). Transfer 107, 445 0985).
228. LEE, K. Y. and TIE~, C. L., Combust. Sci. Technol. 43, 258. GORE, J. P., JENG, S.-M. and FAETH, G. M., J. Heat
167 (1985g TransJer 109, 165 (1987).
229. LEE, K. Y. and TIEN, C. L., Int. J. Heat Mass TransJer 259. GORE, J. P., JENG, S.-M. and FAETH, G. M., Spectral
29, 1237 (1986). and total radiation properties of turbulent carbon
230. ARPACI,V. S. and TABACZYNSKI,R. J., Combust. Flame monoxide/air diffusion flames, AIAA Pap. No. 86-
57, 169 (1984). 0294 (1986).
231. XJEU,D. V., MASUDA,T., COGOLLJ. G. and ESSENHIGH, 260. JENG, S.-M. and FAE'n-I,G. M., J. Heat TransJer 106,
R. H., Eighteenth Symposium (International) on Com- 886 (1984).
bustion, pp. 1461-1469, The Combustion Institute, 261. GORE, J. P. and FAETH,G. M., Twenty First Symposium
Pittsburgh, PA (1980). (International) on Combustion, The Combustion Insti-
232. CO<~OLI, J. G. and ESSENHIGH, R. H., Combust. Sci tute, Pittsburgh (in press).
Teehnol. 16, 177 (1977). 262. ECHIGO, R., Heat Transfer--1982, Vol. VI, pp. 361-
233. KREZINSKI,J. R., BUCKIUS, R. V. and KRIER, H., Prog. 366, Hemisphere Publishing Corp., Washington, D.C.
Energy Combust Sei. 5, 31 (1979). (1982).
234. TsuJi, H., Proo. Energy Combust. Sei. 8, 93 (1982). 263. ECHIGO, R., High Temperature Equipment, A. E.
235. LAW, C. K., Prog. Energy Combust. Sei. 10, 29 (1984). Sheidlin (Ed.), pp. 41-72, Hemisphere Publishing
236. LIU, C. N. and SHIH, T. M., J. Heat Transfer 102, 724 Corp., Washington, D.C. (1986).
(1980). 264. ECHIGO, R., KURUSU, M., ICHIMIYA, K. and
237. SInULraN, M., KULKARNI, A. K. and ANNAMALA1,K., YOSHIZAWA, Y., A S M E - J S M E Thermal Engineering
Eighteenth Symposium {International) on Combustion, Joint Con/erenee, Y. Mori and W.-J. Yang (Eds), Vol.
pp. 611-617, The Combustion Institute, Pittsburgh IV, pp. 99-103, ASM E-JSME, New York (1983).
(1981). 265. ECHIGO, R., HANAMURA, K., YOSHlZAWA, Y. and
238. BIER, R. A., PAGNI, P. J. and OKO8, C. I., Combust. Sci. TOMIMURA,T.. Proceedings hlternational Symposium on
Teehnol. 39, 235 (1984). Heat Trans/er, Vol. 3, pp. 186-193, Beijing (1985).
239. DERIs, J., Seventeenth Symposium (International) on 266. YOSHIZAWA. Y., ECHIGO, R., TOMIMURA, T.,
Combustion, pp. 1003-1016, The Combustion Institute, HANAMURA, K. and KODA. M., Trans. J.S.M.E., B-51,
Pittsburgh (1979). 466 ~1985).
240. CALCOTTE,H. F., Combust. Flame 42, 215 (1981). 267. ECHIGO, R., Proceedings of the US-Japan Joint Heat
241. JENG, S-M., LAI, M-C. and FAETH,G. M., Combust Sei. TransJi, r Seminar, San Diego, CA (Sept. 1985).
Technol. 40, 41 (1984). 268. ECHIGO, R., YOSHIZAWA, Y., HANAMURA, K. and
242. BIER, P. A. and PAGNI, P. J., J. Heat Transfer 105, 109 TOMIMURA,T., Heat Transfer--1986, C. L. Tien, V. P.
(1983). Carry and J. K. Ferrell (Eds), Vol. 2, pp. 827-832,
243. RAMACHANDRA,A. and RAJUNANDAN,B. N., Comhust. Hemisphere Publishing Corp., Washington, D.C.
Sci. Technol. 38, 59 (1984). (1986).
244. DIXON-LEWlS, G., DAVID, T., GASKELL, P. H. et al., 269. GOSMAN, A. D., LOCKWOOD, F. C., MEGAHED, I. E. A.
Twentieth Symposium (International) on Combustion, and SHAH, N. G., J. Energy6, 353 (1982).
pp. 1893-1904, The Combustion Institute, Pittsburgh 270. BoYSON, F., AYERS, W. H., SWITHENBANK,J. and PAN,
(1984). Z., J. Energy 6, 368 (1982).
245. HAMINS, A. and SESHADlU, K., Twentieth Symposium 271. ROBINSOn,J. F., J. Inst. Fuel58, 116 (1985).
(International) on Combustion, pp. 1905-1913, The 272. RICHTER, W., Scale-up and advanced performance
Combustion Institute, Pittsburgh 0984). analysis of boiler combustion chambers, ASME Pap.
246. NEGRELLI,O. E., LLOYD, J. R. and NOVOTNY, J. L., J. No. 85-WA/HT-80, ASME, New York (1985).
Heat Transfi,r 99, 212 (1977). 273. ENOMOTO,H., ESSENHIGH, R. H. and TSAI, Y.-W., Heat
Radiation heat transfer 159

transfer in a continuous model furnace: A comparison Pittsburgh, PA (1984).


of theory and experiment, ASME Pap. No. 75-HT-5, 304. TRUELOVE, J. S., Twentieth Symposium (International)
ASME, New York (1975). on Combustion, pp. 523-530, The Combustion Insti-
274. Gr~RMt~RDONK,R., Glastech. Ber. 36, 886 (1963). tute, Pittsburgh (1984).
275. LEBEDEV, V. I. and SOKOLOV, V. A., Glass Ceram. 33, 305. LEFEBVRE, A. H., Gas Turbine Combustion, Chap. 8,
352 (1976). Hemisphere Publishing Carp, Washington, D.C.
276. LEBEDEV,V. I. and SOKOLOV,V. A., Glass Ceram. 3% 6~ (1983).
(1980). 306. ME~GiJg:, M. P., CUMMINGS,W. G., lit and VISKANTA,
277. HOT'rEL, H. C., J. Inst. Fuel 34, 220 (1961). R., J. Propul. Energy 2, 241 (1986).
278. HOTTEL, H. C., SAROnM, A. F. and FARAG, I. H., 307. NAJJER, Y. S. H. and GOODYEAR, E. M., J. Heat
Combustion Technology: Some Modern Developments, Transfi,r 105, 82 (1983).
Chap. VII, Academic Press, New York (1974). 308. SWlTI-IENBANK,J., TURAN, A. and FELTON, P. G., Gas
279. BEER, J. M., Combustion Technology: Some Modern Turbine Combustor Design Problems, A. H. Lefebvre
Det'elopments, Chap. VIII, Academic Press, New York (Ed.), pp. 249-314, Hemisphere Publishing Corp.,
(1974). Washington, D.C. (1980).
280. ESS~HIGH, R. H., THEKDL A. C., MAt~OUTRE, G. and 309. EWAN, B. C. R., BOYSON, F. and SWITHENBANK, J.,
TSAI, Y., Combustion Technology: Some Modern Twentieth Symposium (International) on Combustion,
Developments, Chap. XIII, Academic Press, New York pp. 541-547, The Combustion Institute, Pittsburgh
(1974). (1984).
281. HOTTEL, H. C., Heat TransJer in Flames, N. H. Afgan 310. CHANG, S. L. and RHEE, K. T., Computation of
and J. M. Beer (Eds), pp. 5-28, Scripta Book Co., radiation heat transfer in diesel combustion, SA E Pap.
Washington, D.C. (1974). No. 831332 (1983).
282. JOHNSON,T. R. and BEER, J. M., Fourteenth Symposium 311. CHANG, S. L., YANG, X. L. and RrIEE, K. T.,
(International) on Combustion, pp. 639-649, The Com- Proceedings on the InteJ~ational Symposium on Diag-
bustion Institute, Pittsburgh, PA (1973). nostics and Modeling of Combustion in Reciprocating
283. STEWARD, F. R., OStJWAN, S. and PlCOT, J. J. C., Engines, JSME/SAE/MESJ, Tokyo, Sept. 4-6 (1985).
Fourteenth Symposium (International) on Combustion, 312. CHAPMAN, M., FRIEDMAN, M. C. and AGI-IAN, A., A
pp. 651-660, The Combustion Institute, Pittsburgh, time-dependent spatial model for radiant heat transfer
PA (1973~ in diesel engines, SAE Pap. No. 831725 (1983).
284. MICHELFELDER,S. and LOWES, T., Glasteeh. Ber. 46, 99 313. MENt~OtT, M. P., VISKANTA, R. and FEROUSON, C. R.,
(1973g Multidimensional modeling of radiative heat transfer
285. JESCHAR, R. and SCHUPE, E., Gas Win'me lntl 22, 473 in diesel engines, SAE Pap. No. 850503 (1985) (also in
(1973). SAE Trans.--1985).
286. GUNTHER, R., JESCHAR, R., POTKE, W. and SCHUPE, W., 314. WHITEHOUSE,N. D. and SHAHAD,H. A. K., Flows in
Gas Warme lntl 23, 381 (1974). Internal Combustion Engines, T. Uzkan (Ed.), ASME,
287. SCHUPE, W. and JESCHAR, R., Gas Warme Intl 24, 64 New York (1984).
(1975). 315. ANNONYMOUS, A review of compartment models,
288. MICHELFELDER,S., Glastech. Ber. 50, 193 (1977). National Bureau of Standards, Rep. BSIR 83-2684
289. DETKOV,S. P., J. Engng Phys. 46, 464 (1984). (1983).
290. HAOVIG,S., J. Inst. Fuel 35, 129 (1970). 316. MUDAN, K. S., Prog. Energy Combust. Sci., 10, 59
291. HADVIG,S., J. Inst. Fuel 35, 202 (1970). (1984).
292. FRISCH, V., JESCHAR, R. and POTKE, W., Gas Warme 317. YANG, K. T. and LLOYD, J. R., Natural Convection:
Intl 31,421 (1982). Fundamentals and Applications, S. Kakac, W. Aung
293. FmSCH, V., JESCHAR, R. and POTKE, W., Gas Warme and R. Viskanta (Eds), pp. 303-329, Hemisphere
Intl 31,476 (1982). Publishing Corp., Washington, D.C. (1985).
294. HUTCHINSON,P., KHALIL, E. E. and WHITELAW,J. H., 318. YANG, K. T. and LLOYD, J. R., Natural Convection:
Turbulent Combustion, L. A. Kennedy (Ed.), pp. Fundamentals and Applications, S. Kakac, W. Aung
211-228, AIAA, New York (1978). and R. Viskanta (Eds), pp. 381-410, Hemisphere
295. GOSMAN,A. D., LOCKWOOD, F. C. and SALOOJA,A. P., Publishing Corp., Washington, D.C. (1985).
Seventeenth Symposium (International) on Combustion, 319. YANG, K. T., Heat TransJer--1986, C. L. Tien, V. P.
pp. 747-760, The Combustion Institute, Pittsburgh Carey and J. K. Ferrell (Eds), Vol. 6, pp. 131-140,
(1979). Hemisphere Publishing Corp., Washington, D.C.
296. ABou ELLAIL, M. M. M., GOSMAN, A. D., LOCKWOOD, (1986).
F. C. and MEGAHED, I. E. A., Turbulent Combustion, L. 320. COOPER, L. Y., Nineteenth Symposium (International)
A. Kennedy (Ed.), pp. 163-190, AIAA, New York on Combustion, pp. 933-939, The Combustion Insti-
(1978). tute, Pittsburgh, PA (1982).
297. DOLEZAL, R., Large Boiler Furnaces, Elsevier, 321. COOPER, L. Y., Twentieth Symposium ~lnternational) on
Amsterdam (1967). Combustion, pp. 1567-1573, The Combustion Institute,
298. FIVELAND, W. A. and WESSEL, R. A., FURMO: A Pittsburgh, PA (1984).
numerical model for predicting performance of three- 322. COOPER, L. Y., Combust. Sci. Technol. 40, 19 (1984).
dimensional pulverized-fuel fired furnaces, ASME 323. BAGNARO, M., LAOU1SSET, M. and LOCKWOOD, F. C.,
Pup. No. 86-HT-35 (1986). Fire Dynamics and Heat Transfl,r, J. G. Quintiere, R. L.
299. GOODRICH, A. M. and READ, A. W., Pray. Energy Alpert and R. A. Altenkirch (Eds), pp. 107-117, ASME,
Combust. Sei. 2, 83 (1976). New York (1983).
300. LowE, A., WALL, T. F. and STEWART, I. McC., 324. MARKATOS, N. C. and PEmCLEOUS, K. A., Fire
A.I.Ch.E JI 23, 440 (1977). Dynamics and Heat Transfi,r, J. G. Quintiere, R. L.
30l. LOCKWOOD, F. C., SALOOJA, A. P. and SYED, S. A., Alpert and R. A. Altenkirch (Eds), pp. 115-124, ASME,
Combust. Flame 38, 1 (1980). New York (1983~
302. LOCKWOOD,F. C. and SALOOJA,A. P., Combust. Flame 325. JOULAIN, P., MOST, J. M., FUSEAU, Y. and SZTAL, B.,
54, 23 (1983). Seventeenth Symposium (International) on Combustion,
303. LOCKWOOD, F. C., RIzvL S. M. A., LEE, G. K. and pp. 1041-1051, The Combustion Institute, Pittsburgh,
WHALEY, H., Twentieth Symposium (International)on PA (1978).
Combustion, pp. 513-522, The Combustion Institute, 326. BRZUSTOWSKI,T. A. and TWARDUS, E. M., Nineteenth
160 R. VISKANTAand M. P. MENG0~

Symposium (International) on Combustion, pp. 847-854, 330. ALBINI,F. A., Comhust. Sci. Technol. 42, 229 (1985).
The Combustion Institute, Pittsburgh, PA (1982). 331. ALBINI, F. A., Combust. Sci. Technol. 45, 101 (1986).
327. HWANG, L. C. and LITTON, C. D., Twentieth Sym- 332. BROSMER,M. A. and TIEN, C. L., J. quantve Spectrosc.
posium (International) on Combustion, pp. 1673-1679, radiat. TransJt, r 33, 521 (1985).
The Combustion Institute, Pittsburgh, PA (1984). 333. BROSMER, M. A. and TILm, C. L., Heat TransJer 107,
328. WICHMAN,I. S., Combust. Sci. Technol. 40, 233 (1984). 943 (1985).
329. HAYNES, B. S. and WAGNER, H. G., Prog. Energy 334. BROSMER,M. A. and TIEN, C. L., Combust. Sci. Technol.
Combust. Sci. 7, 229 (1981). 48, 163 (1986).