Pergamon

PII: SOOl7-9310(96)00139-l

Evaluation of discrete ordinates method for

radiative transfer in rectangular furnaces
NEVIN SELCUK and NURAY KAYAKOL
Department of Chemical Engineering, Middle East Technical IJniversity. Ankara 06531. Turkey

(Receiced 25 Octohw 1995 md in firm/ form IO April 1996)

Abstract-The discrete ordinates method (DOM) and discrete transfer method (DTM) were evaluated
from the viewpoints of both predictive accuracy and computational economy by comparing their pre-
dictions with exact solutions available from a box-shaped enclosure problem with steep temperature
gradients. Comparative testing shows that the S, approximation produces better accuracy in radiative
energy source term than in flux density in three orders of magmtude less CPU time than that required by
the DTM. The S, approximation can therefore be recommended to be used in conjunction with CFD
codes. Copyright $~) 1996 Elsevier Science Ltd.

1. INTRODUCTION DTM (discrete transfer method) which combines vir-

tues of the first three. The Hottel’s zone method and
The ever-increasing consumption of limited energy the Monte-Carlo technique have long been accepted
sources on one hand and environmental concern over as the most accurate methods for calculating radiative
the reduction of combustion-generated pollutant transfer. These traditional methods, however, have
emission on the other, necessitates improvement of not found application in comprehensive combustion
thermal and emission performances of combustion modeling due to their large computational time and
systems burning natural gas, oil and coal. This can be storage capacity. At the same time. neither of these
achieved by developing mathematical models which models requiring the solution of simultaneous
could eventually be used for the design of environ- algebraic equations are compatible with the sohltions
mentally acceptable combustion systems with of the differential equations governing the transport
improved energy efficiency. Thermal radiation is the of mass, momentum, species and energy in com-
predominant mode of heat transfer at the high tem- prehensive numerical models.
peratures encountered in most industrial combustion The so-called flux methods which provide an alter-
chambers. Therefore, an adequate treatment of ther- native but less accurate method of solution convert
mal radiation is essential to develop a mathematical the integro-differential equation into differential equa-
model of combustion systems. tions by discretization of angular variation of radi-
The exact (analytical or numerical) solution of inte- ative intensity [557]. Being differential in form these
gro-differential radiative transfer equation (RTE) is equations are ideally suited to numerical solution sim-
generally a formidable task. Although there have been ultaneously with the flow equations. Therefore, the
a few attempts to formulate RTE for non-isothermal flux methods offer a very high degree of computational
rectangular enclosures [I, 21, explicit solutions are economy and for this reason they have been widely
only available for simplified situations such as black employed in combustion models. All conventional
walls and constant properties etc. [3, 41. Nevertheless flux methods suffer from the difficulty in application
exact solutions even for these simplest systems are to complex geometries whereas some of them suffer
needed, as they can serve as benchmarks against which from an inadequate number of directional fluxes and
the accuracy of approximate solutions is tested. a lack of coupling between these fluxes. The DOM
There is growing interest in approximate multi- (discrete ordinates method), which conceptually is an
dimensional radiative heat transfer models mainly extension of flux methods, corrects these defects. The
in response to the need for accurate, flexible and com- DOM achieves a solution by solving the exact RTE
putationally economic models which couple easily for a set of discrete directions spanning the angle range
with the differential equations governing the transport of 47r. Apart from the obvious ease with which the
of mass, momentum, species and energy in com- method can be incorporated into computational fluid
prehensive models for turbulent, reacting and rad- dynamics (CFD) calculations, it is relatively easy to
iating flows. code and requires single formulations to invoke higher
The available approximate solution methods can order approximation. This method was first suggested
be classified into four groups : Hottel’s zone method, by Chandrasekhar [S] for one-dimensional astro-
Monte-Carlo methods, flux-type methods and the physics problems. More recently. Carlson and

213
214 N. SELCUK and N. KAYAKOL

NOMENCLATURE

A, B, C control volume face areas in I, y, z Subscripts

direction [m’] burner burner wall
d slope of the axial temperature curve bw black wall
E emissive power [W m--‘I e end-face
I radiant intensity [W rn-’ sr’] end end wall
k, gas absorption coefficient [mm’] g gas
L half of the side lengths of the i inlet
rectangular enclosure [m] m discrete direction
f direction cosines max maximum
4 radiative heat flux [W m-‘1 0 reference value
9net net radiative heat flux [W mm’] P control volume centre
r position vector r reference-face
s radiative energy source term [W mm’] side side wall
T temperature [K] W wall.
uP volume of pth control volume [m’]
M’, weight function in a direction m
Superscripts
“,Y, -7 co-ordinate axes in Cartesian
+ leaving
geometry _
z’ arriving
dimensionless axial distance.
dimensionless.

Greek symbols
finite-difference weighting factor Abbreviations
: polar angle CFD computational fluid dynamics
p, 9, 5 direction cosines in X, y, z directions DOM discrete ordinates method
g Stefan-Boltzmann constant DTM discrete transfer method
7 optical thickness DT64 discrete transfer using 64 ray per wall
f$ azimuthal angle node
51 direction vector of radiant intensity. RTE radiative transfer equation.

Lathrop [9] developed the DOM to solve multi-dimen- (2) The accuracy of DOM has been tested in iso-
sional neutron transport problems. Over the past lation from the modeling of other physical processes.
decade, the method has been applied to one- [lo, 1 l] Testing has taken two forms : (i) the model has been
and multi-dimensional [12-211 radiative heat transfer employed by using prescribed radiative energy source
problems. term distributions (zero [14, 151 or uniform [14, 16,
Previously published multi-dimensional evalu- 171) and comparing the predicted temperature and
ations of the accuracy of the DOM have taken three radiative heat flux distributions with the Hottel’s zone
forms : method or exact solutions. This procedure for the
(1) The DOM has been employed as part of a com- evaluation of the accuracy of the DOM suffers from
plete prediction procedure: (i) the predicted tem-
two major disadvantages : (a) even if acceptably accu-
perature distributions have been compared with rate predictions are obtained for the uniform source
experimentally determined data [ 121. Discrepancies term distributions, there is no certainty that similarly
between predicted and measured values may be partly accurate predictions will be produced for the highly
due to the errors in the experimentally determined non-uniform distributions encountered in operating
data ; even if the experimentally determined data are furnaces and combustors ; (b) considering the iterative
correct, it is impossible to decide whether dis- sequence of the solution in the complete prediction
crepancies in the predicted temperature and radiative procedure, input data provided should be temperature
heat flux distributions are attributable to the radiation distributions, and the predicted and tested quantities
model employed, or to inaccuracies in the models should be radiative heat flux and radiative energy
used for the prediction of flow, reaction, etc. (ii) The source term distributions. (ii) The model has been
predicted temperature and heat flux distributions have tested by using prescribed temperature distributions
been compared with those predicted by the six-flux (uniform [ 15, 16, IS] or non-uniform [ 16, 17, 191) and
model [ 131. However, assessment has been carried out comparing the predicted heat flux distributions with
on a small laboratory furnace where radiation was those predicted by the Hottel’s zone method. This
not the controlling heat transfer mechanism. procedure has two drawbacks : (a) although uniform
 Radiative transfer in rectangular furnaces 215

temperature distributions are computationally con-

venient they do not even approximately represent the
extremely non-uniform distributions encountered in
operating furnaces and combustors; (b) the use of
Hottel’s zone method for testing the radiative heat

..$yy
;
flux distributions of an enclosure problem with steep
temperature gradients requires finer zoning, which can
be computationally inefficient due to extensive
x>P___....._..__._...................
___............
machine storage capacity and time. 8 .,.-- I(r,lR)
(3) The DOM has been tested by calculating wail . .::. :
. ‘. _:

heat flux distribution from measured temperature dis-

..............__
.-.__
4::
......2 =,5
tribution and comparing the predictions with
measurements [I 3,201. This assessment procedure can YVq
be misleading due to the decoupling between the radi-
ation model and the models used for the prediction of Fig. I Coordinate system.

flow, reaction, etc.

What is required at the present time is testing of

the accuracy and computational economy of the two fl, (p,,, = cos 8, Q,, = sin 0 sin 4, 5, = sin 8 cos b), k,
radiation models (namely the DOM and the DTM) absorption coefficient of medium, Z,, the total black-
extensively used as part of complete prediction pro- body radiation intensity at the temperature of the
cedures in the past decade, in isolation from the mod- medium. If the surface bounding the medium is
els of flow and reaction, on problems with highly assumed to be black, then the radiative boundary
non-uniform gas temperature distribution typical of condition for equation (1) is given by
operating furnaces. Assessment of the predictive accu-
racy of the DTM has been reported in detail in an I” = Ibw = aT: /TT (2)
earlier paper [21].
In the present paper, the DOM is applied to solve
where Zbwis the blackbody radiation intensity at the
the radiative heat transfer problems in two test cases
temperature of the surface, T,“. Equations (1) and (2)
for which analytical and numerical exact solutions are
represent m coupled partial differential equations for
available. The first is a one-dimensional slab con-
the m intensities, Z”‘.
taining an absorbing-emitting medium with uniform
The finite-difference form of equation (1) can be
gas temperature [22], the second is a box-shaped
obtained by multiplying equation (1) by dx* dyed:
enclosure problem based on data reported previously
and integrating over the control volume as follows
on a large-scale experimental furnace with steep tem-
perature gradients. The test cases are chosen to bench-
mark the DOM against exact solutions [3] and the
other most widely used model, namely the DTM. The
DOM and the DTM are also evaluated from the view-
point of computational economy.
where I”’ is cell-center intensity and
I’& I’$, Is, Jg, I;“,, Z$ are the intensities at each face of
2. DESCRIPTION OF THE METHOD control volume. If the face intensities I’$ I:, and I:;
are assumed to be known from boundary conditions
The physical situation to be considered is that of
for a control volume adjacent to a boundary of the
an enclosure containing an absorbing-emitting, non-
enclosure, the following relationships [14] can be used
scattering radiatively gray medium whose absorption
to eliminate the unknown face intensities ZTe,Zz and
coefficient is the same at all points, and of known
Z$ in equation (3).
magnitude. Values of black-body emissive power are
assumed to be available at all points within the
enclosed medium and at all points on the interior r; = aZ:,+(l-z)z: (4)
bounding surfaces of the enclosure which are assumed
to be black. Iz’ = aZ”,+(l -cY)ZF (5)
The discrete ordinates vepresentation of RTE for
an absorbing-emitting gray medium in a rectangular r; = rz;+(l -r)ZZ. (6)
coordinate system (see Fig. 1) can be written as :
The finite-difference weighting factor c( = 0.5 rep-
resents the second-order diamond difference scheme
proposed by Carlson and Lathrop [9]. After some
where I” [=1(.x, y, z ; p,,,,Q,,,&,,)I is the total radiation rearrangement the cell-center intensity, Z; may be
intensity at position (x,_!J,z) in the discrete direction evaluated as
216 N. SELCLJK and N. KAYAKOL

rature scheme. Schemes for one-dimensional radiative

transfer problems are well developed but their accu-
racies are significantly dependent on the physical nat-
where ure of the problem. However, the extension of these
A = A_rAz, B = AxA:, c‘ = A.YAJ. angular quadrature schemes to multi-dimensional
(8)
radiative transfer problems is not always obvious and
Every I”’ value on every cell-cenrer and faces of straightforward. The choice of scheme is arbitrary
control volume can be computed by stepping from although restrictions on the directions and weights
control volume to control volume. The direction of arise from the need to preserve symmetries and invari-
recursive evaluation is in accord with the direction of ance properties of the physical system [9].
physical propagation of radiation beam as defined by
(Pmrul,111
id.
” ’
Once the intensity distribution is determined, quan- 3. PERFORMANCE OF DISCRETE ORDINATES
tities of interest such as radiative flux and energy METHOD
source term distributions can be readily evaluated.
The DOM has been evaluated on two test cases for
The net radiation heat flux on a wall is calculated as
which analytical and numerical exact solutions are
4 available. The first being a one-dimensional slab con-
(!llCI= 4 -Y c4
taining an absorbing-emrtting medium with uniform
where q’ and y are leaving and arriving wall heat gas temperature [22], the second being a box-shaped
Auxes, respectively. For the physical situation under enclosure problem based on data reported previously
consideration, CJ+and q- can be written as on a large-scale experimental furnace with steep rem-
perature gradients [3].
(io)

3.1. A one-dimmswnul shb conraming un absorbing-

emitting medium M.ith umjorm gas rrmprrature
The physical srtuatton is two infinitely large parallel
The radiative energy source term distribution for
cold black plates containing an absorbinggemitting
problems where prescribed temperature distributions
medium with uniform temperature. Wall heat Ruxes
are available is expressed as
calculated by using DOM for varying optrcal thick-
nesses between 0.1 and 3.0 and for different angular
quadrature schrmes,were compared with the exact
anaiytrcal soiutrons 1221.
As can be seen born equations (11) and (IL?),yudd- Figure 2 shows the comparison between exact solu-
rature schemes are required for approximating the tions and DOM predictions using two different angu-
integrals. Therefore the accuracy of discrete ordinates lar quadrature schemes ; S,, (S2 and A’, [15]) and S,,,
solutions depends on the choice of the anguiar quad- (S, and S4 [IS]). As shown in Fig. 2, different angular

1.2 I_ -
_______------ ______-_--
1 .1 __-- --
,*--
: _.-.-.,-
1 .cl 1 /’ ~..;.,.;:.A:-.
_),___ ,.-... _:_. .

I’ _I
,_...
_.-’ ,,__. .’
20 0.9 -
,’
,
:_/.
_’
,:.
,_
_.C ,_ ”

‘a ,’ .’
^ 0.6 -
I’.. /
: ,/

: ,‘,.:/;” - Exact
= 0.7 4 .,: s*
z I

/.‘/.
.,

._..._, s4
J. -:
2 0.6 - ,
J
.,
//
---___ s*
$ /:t

P 0.5 - ,rj7 .I._._ c&

E
=S 0.4 -
$ :
f
6E 0.3 -
r
0.2 1 /

0.1 I r i i
0.0 0.5 1 .o 1.5 2.0 2.5 3.0
Optical tmckness, 2
Fig. 2. Con~parwn berwernrxac~solutrons and UOM predictions of‘dirnensronless heat nuxes.
 \
Radiative transfer m rectangular furnaces 217

1 .-.,-_
roF low orders of G-N do not predict the half range heat
9 ‘. fluxes accurately as they do not correctly integrate the
..-- Q+

6 ‘\\ - F-N half range first-order moment of intensity. To achieve

‘.. . .-.. 6-N
reasonable accuracy, higher order approximations of
\ ‘..
-. \
1
G-N are necessary at the expense of computational

I_:::
6 ’
‘..
‘-__
--__ inefficiency unless S-N or F-N schemes are employed.
\ --__
5
7 -_ Wall heat fluxes predicted by both S-N and F-N com-
--__
4 ‘-.
‘A pare well with the exact solution as they both satisfy
“,
3
‘b._._
the first-moment for half range. It can be concluded
2 that for the test problem under consideration the half
IL range integrals in the boundary conditions are more
2 4 6 6 important to solution accuracy than the full range
Order OfIIOM integrals in the R’I‘E.
Figure 4 indicates the variation of absolute per-
centage errors in wall heat fluxes predicted by the
22 i
DTM with number of rays for various optical thick-
nesses. As can be seen from Fig. 4, the accuracy
increases with increasing ray numoers and only slight
improvement is obtained after a certain number of
rays.

3.2. A three-dimmslonul rectunyulur enclosure prob-

lrm contuimng un absorbing-emitting medium with
steep temperature gradients
The accuracy of DOM for three-dimensional radi-
ative heat transfer has been assessed by applying this
2 4 6 6
radiation model to the prediction of distributions of
OnYerotDOM
radiative flux density and radiattve energy source term
.”
of a highly demanding rectangular enclosure problem
c with a steep temperature gradient and a non-scattering
16 -\
. \ medium of small optical thickness and by comparing
14- ‘\
\ its predictlons with exact solutions produced pre-
12- \
\ viously [3]. The rectangular enciosure under con-
10 - \
\ sideration has interior black walls and an absorbing-
\
II- \ emitting medium of constant propernes. ‘The problem
\
6- \ wa5 based on data taken from a large-scale cxper-
\
\ imental furnace with steep temperature gradients typi-
4- --W_
_--_ cal of operatmg furnaces.
2- ---_ -----____
I’he experimental furnace under consideration is
I horizontal, oftunnel type with a square cross-section,
4 6 6 it is fired horizontally from the center of the left end
Order at UOM
wall, which is the burner wall, with a rmxture of oil
Fig. 3. Variation of solution errors with the order for S-N. and gas with no swirl, and operates at atmospheric
G-N and F-N angular quadrature schemes.
pressure, the four side walls are water cooled, and the
burner and back end walls are refractory. Table 1
shows the complete dirnenstonless data obtained from
quadrature schemes (i.e. the selection ofdiscrete direc- the expernnental furnace and used as input data for
tions and associated weights) may result in con- both the exact solution and DOM. A detailed descrip-
siderably different accuracies. S,, significantiy deviates tion of data can be found elsewhere [3].
from the exact solution. .S, produces better agreement The numerical solution procedure in the present
than S,, due to the failure of S, to satisfy the half work has been based on the method of Carlson and
range flux conditions [ 151. Lathrop [9]. R’l’E has been reduced to conservative
Figure 3 illustrates variation of absolute percentage finite difference form using the control volume method
errors in wall heat fluxes with the order of DOM using and the diamond d:fference scheme with set-to-Lero
S,, (S-N), Gaussian (G-N) and Fiveland (F-N) [lo] negative intensity fix-up. Quadrature pomts and
angular quadrature schemes for various optical thick- weights are calculated by the moment-matching tech-
nesses. As can be seen from Fig. 3, for all optical nique proposed by Carlson and Lathrop 191. fhe
thicknesses, percentage errors in wall heat fluxes quadrature scheme satisfies the zeroth and second
decrease with the order of DOM, irrespective of the moments for full rdnge (i.e. 4~) ; and the first moment
angular quadrature schemes. It can also be noted that for half range (i.e. 2x). The S,, discrete ordinate for-
218 N. SELCUK and N. KAYAKOL

3.0

10 20 30 40 50 60 70 80 90 1’00
Number of rays
Fig. 4. Variation of solution errors with number of rays for DTM.

Table I. Dimensionless data fed to the exact solution and S, dictions of the radiative flux density and energy source
method program
term of the same enclosure problem. Comparisons of
Dimensions of the furnace L”, = 1, L; = 1, E: = 6 the DTM predictions employing different ray numbers
and control volumes with the exact solutions indicated
Optical thickness 7” = l/6
that the DTM provides radiative flux density and
Wall black-body intensities (Q,,& = 0.0020 energy source term distributions in close agreement
(&Jburner = 0.0574 with the exact solutions for 64 rays and for the same
(I&,, = 0.0167
subdivision of the enclosure. For comparative testing
Gas temperatures ? = 0.1775 of the two methods from the viewpoint of both the
<= = 0.6222
accuracy and computational economy the predictions
T Ill&= 1
of both will be presented here.
Position of the peak z’,,, = 0.8 3.2.1. Flux density distributions. Figure 5 illustrates
Slope of gas temperature de = -0.220 the comparison between the point values of the dimen-
distribution at furnace exit sionless heat flux density to the side wall in the positive
x-direction predicted by the S, approximations, the
Reference values used to make the experimental data dimen-
DT64 and exact solutions for surface grid points.
sionless: L, = 0.48 m, T,, = 1673K, & = 4.4419 x 10’ W
mm’. lo = 1.4139 x lo5 W mm2 sr-‘.
Points on the lines (a = 1,j = 0.25,F) and
(-i- = 1, j = 0.75, i) represent points near the center of
the face and near the corner of the face, respectively.
mulation requires n(n + 2) discrete directions for As can be seen from Fig. 5, the flux densities to the
n = 2, 4, 6,. . The number of directions required in side wall predicted by the S, are found to be in good
a completely symmetric arrangement is found to be agreement with the exact solutions over the whole
n(n+2)/8, i.e. the total number of directions per length of the enclosure. A condensed comparison of
octant. The quadrature ordinates and weights for the the S, and the DT64 predictions of flux densities is
S, approximation are listed in Table 2. The test enclos- contained in Table 3. As can be seen from Table 3,
ure has been subdivided into 4 x 4 x 24 control vol- both methods produce acceptable accuracies although
umes in x, y, z directions, respectively. In an earlier the average absolute error produced by the S4 is two
paper [21], the DTM had been applied to the pre- times higher than that produced by the DT64.

Table 2. S, angular quadrature scheme the first octant

Direction Ordinates Weight

number Ilm q, h’,

1 0.9082483 0.2958759 0.2958759 0.52359X7

2 0.2958759 0.9082483 0.2958759 0.5235987
3 0.2958759 0.2958759 0.9082483 0.5235987
 Radiative transfer in rectangular furnaces 219

0.10 1

- Exact
._.-.- f)T64

0.03 -

~02~““‘~‘~“‘~““““““’
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Dimensionless furnace length, ?
Fig. 5. Comparison between the exact values and the DTM and the S, predictions of dimensionless flux
densities to the side wall.

Table 3. Comparison of the S, and the DT64 predictions of dimensionless flux densities for
surface grid points

Average absolute error Maximum

Method (2 = 1,~ = 0.75,s) (.f = l,?; = 0.25,?) Side wall percentage error
___~
DT64 1.12 1.93 1s2 10.21
& 3.40 2.87 3.13 9.43

0.40

0.35
- Exact
.-.-.- DT64
10 0.30 ._..._. S4

E-
&I
- 0.25
g
S
$ 0.20
I
S!
s 0.15
._
t
g 010
ii
0.05

0.00 I, / I , I , , / , I, I, / / , I I /
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Dimensionless furnace length,?
Fig. 6. Comparison between the exact values and the DTM and the .S, predictions of dimensionless radiative
energy source terms along (.u”= 0.25,9 = 0.25. I).
220 N. SELCUK and N. KAYAKOL

0.12

0.11 - - Exact
.-.-.- D-r64
zu
E^
& 0.10
ar
2
: 0.09 -

f
Z
.o 0.08 -
t
2
ii 0.07 -

0.06 ‘. ’ ’ ’ ’ / I 1 *., I, / I ,

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Dimensionless furnace length, “r
Fig. 7. Comparison between the exact values and the DTM and the S, predictions of dimensionless EidiatiVe
energy source terms along (1 = 0.75.y = 0.25,n.

t 0.09 /’

1_-L
a, /’

z
._
__._._._.-.-.-.J_L:‘_~..._..
_._._,_.-.-‘-.
e 0.06
‘.. -._,_._._.-.-.
,._.,._...-.”
_,_._’

E ..‘--.‘-. ._,,,_,,,
,,_.----~
ii
0.07 .

0.06 8, I,I~I,,,!,.,/ 1 /

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Dimensionless furnace length,?
Fig. 8. Compartson between the exact values and the DTM and the S, predictions of dimensionless
radiative energy source terms along (.Z = 0.75, F = 0.75.9).

Table 4. Comparison of the S, and the DT64 predictions of dimenstonless source term disrrtbutions
for medium grid points

.4verage absolute error Maxrmum

Method (a = 0.75,~ = 0.75,?) (.f = 0.75.V = 0.25,?) (1 = 0.25.y = 0.25,Z) percentage error

DT64 9.36 2.27 1.35 11.64

& 1.75 0.62 0.63 3.66
 Radiative transfer in rectangular furnaces 221

Table 5. Comparisons of the S, and the DT64 predictions of dictions of the distributions of radiative flux density
the percentage errors in generared and removed radiative and energy source term of a black-walled rectangular
energy
_____ _ _~ ~ enclosure problem. The problem was based on data
Percentage error Percentage error taken from a large-scale experimental furnace with
Method in generation in removal steep temperature gradients typical of operating fur-
naces. The accuracy of the models have been tested by
D-l-64 -2.34 -2.12
comparing their predictions with the exact solutions
‘5’
I 0.29 0.29
reported earlier in the literature. Comparisons show
that the S, approximarlon produces better accuracy
in radiative energy source term than in flux density to
3.2.2. Source term distributions. Figure 6 shows the the side wall. Comparative testing of the S, approxi-
comparison between the S, and the DT64 predictions mation and discrete transfer methods indicate that
of dimensionless source term distributions and exact their predictions are in good agreement with the exact
va!ues for points (.i = 0.25,~~ = 0.25, ?). These grid solutions and that the S, approximation requires a
points represent the points at the cemer of the row of CPU time three orders of magnitude less than that
control volumes near-est to the furnace axis. It can be required by discrete transfer method. Therefore, the
seen that the trend of the distributions predicted by .S, approximation is a better alternative to be used as
the S’, and the DT64 is the same as that of the exact part of a complete prediction procedure.
distriburlons which follow the physically expected
trend, rismg steeply from the burner wall onward.
REFEHENCES
going through a nlaxlmum and decreasing con-
tinuously toward the exit. The maxmlum ofthe source I. P. C’heng. Exact solutions and differenttal approxi-
mation for multi-dimensional radiative transfer in Car-
term distrimnion occurs at the same location as the tesian coordinare connguration, Pray. Asrronuut. Aaw
maximum of temperature distribution. nm~. 31, 260-508 (19723.
Figures 7 and 8 illustrate the comparison between 2. A. L. Crosbie and L. C. Lee, Relation berween multi-
the exact values of dm~ensionless source term and dlmenslonal radialive transfer in cylmdrical and
rectangular coordinales with arusotropic scarrering,
distribuuons predicred by the S, and the DT64 for
J. Quanr. Sp~c~rosc. Radiai. Trunsf~r 38, 331-241
grid points (_\’=T0.7.5? = 0.25, ?) and (.< = 0.75, (lY87).
i; = 0.75, P). These grid pomts represent the medium 3. N. Sel~uk, Exact sotuuons tbr radiattve heat transfer
points nearer to the side wall and near the Conner of in box-shaped furnaces, J. Heat Tronslw 107, 64X-655
the enclosure, respecuvely. A conaensed comparison (1985).
4. N. SelFuk and Z. Tahiroglu. Exact numerical soluttons
of the S,, and the D’J%4 predictions of source term for radiattve hear transfer in cylindrical furnaces. fr!/. J.
distribuuons is contamed in Table 4. As can be seen N~,vrrr. Meth. En,y/lig26, 1201-1212 (IYXX).
from Table 4, the maximum point percentage error 5. N. Selpuk. Evaluarion of flux models for raciianve trans-
produced by the & is approxunately one-third of that fer in rectangular furnaces. Inr. J. Hear Muxs TWIW/P~
31, 3477-1482(lYX8).
produced by the D’P64. It can be noted that the S’,
6. N. SelCuk, Evaluation of flux models for radiative trans-
produces smaller average absolute errors than the fer in cylindrical furnaces. fnt. J. ifrear Muss Trunsfrr 32,
DT64. The CPU times for the S, and the IYr64 using 620-624 ( 1989).
the IBM Kisc/6000 Model SO are 1308 and 2.3 s. 7. N. SelGuk. Evaluation of spherical harmomcs approxi-
respecrively. It can therefore be concluded that the S, mation for radiaGve transfer in cylindrical furnaces. ;w/.
J. Hrut Muss Twn.v/rr 33, 579558 I (I 990).
approxmlauon is not only more accurate but also
8. S. Chandrasrkhar. Nur2iarirr Trunsfbr. Dover, New York
compuratioualiy mom-e efficient than the DT64. (IYhO).
To provide a global check on the accuracy of the 9. B. C. Carison and K. D. Lalhrop. Transport theory--
S4 predlcuoua, the rotai rare of’ removal of radiative the method ofdiscrere ordinates, in Compurrq Merhods
WI Rwcror Phy.sic.s (Edited by H. Greenspan. c‘. N.
energy thtougn walls and the total rate of generation
Kelber and D. Orkent), pp. 165-266. Gordon & Breach.
of radiarlve energy withm the enciosed rnedium have New York (lYh8).
been evaluawi and cornpared with the exact values 10. W. A. Fiveland, Ulscrere orclmate methods for radiative
[3j. Table 5 shows the Edrot’s in generated and rrmoved heat trausrbr in isorruplcaily and amsorroplcally scat-
radlatlve cilergy produced by the S, and the DT64 termg media. J. Hecr/ Trransf>r IUY, X09-8 I2 (1087).
II. J. S. Truelove, An evaluarion of the discrete ordinate5
predictions. As can be seen from Table 5, the per-
approxmtanon for radiatrve transfer in an absorbmg,
centage ertol-s in generated and removed radiative emittmg, and scatrrrmg planar medium. HTFS Report
energy are equal for the S, predictions, but aimost no. KX478 (1976).
equal f’or tile DT64. The discrepancy encountered in 12. E. E. Khalil and J. S. Truelove. Calculation of heat
the DT64 can be cons&red to be due to the evalu- transfer In a large gas fired furnace. H1.F’ Report no.
RX747 (1977).
ation of radiative energy source term through the 13. A. S. Jamaluddin and P. J. Smllh, Predictmg radiauve
method originally suggested by Shah 1221. transfer m axisymrnerric cylindrical enclosures usmg dis-
crete ordinates method. L‘or~zhust. .%i. 7&r10/. 62, 173-
4. CONCLUSIONS IX6 (IYXX).
14. W. A. Fiveland. -1hree-dunensional radiative heat
The S, discrete ordmales approxmlauon and dis- transfer solur~ons by discrete ordinares rnerhod. J. 7%~
crete transfer methods have been applied to the pre- fW,~l,_i’.Y.2, .NY 3 I6 cIYXX).
222 N. SELCUK and N. KAYAKOL

15. J. S. Truelove, Discrete ordinates solutions of radiation 19. A. S. Jamaluddin and P. J. Smith, Discrete-ordinates
transport equation, J. Hear Transfer 109, 1048-1051 solutions of radiative transfer in nonaxisymmetric cyl-
(1987). indrical enclosures, .I. Thermophys. 6,242-245 (1992).
16. A. S. Jamaluddin and P. J. Smith, Predicting radiative 20. W. A. Fiveland, A discrete ordinates method for pre-
transfer in rectangular enclosures using discrete ordi- dicting radiative transfer in axisymmetric enclosures,
nates method, Combust. Sci. Technol. 59, 321-340 ASME paper 83-HT-20 (1982).
(1988). 21. N. SelCuk and N. Kayakol, Evaluation of discrete trans-
17. D. J. Hyde and J. S. Truelove, The discrete ordinate fer model for radiative transfer in combustors. Pro-
approxlmatlon for multi-dimensional radiant heat trans- ceediys oflntwnutionul Symp. on Radiatit’e Heut Truns-
fer in furnaces. HTFS Report no. R8502 (1977). fer. ICHMT. Kusadasi. Turkev (1995).
18. W. A. Fiveland, Discrete ordinates solutions of transport 22. N. G. Shah, New method of comp&ation of radiant
equations for rectangular enclosures, J. Heat Transfer heat transfer in combustion chambers. Ph.D. Thesis.
106,699-706 (1984). University of London, London (1979).